path
stringlengths 9
117
| type
stringclasses 2
values | project
stringclasses 10
values | commit_hash
stringlengths 40
40
| commit_message
stringlengths 1
137
| ground_truth
stringlengths 0
2.74k
| main_code
stringlengths 102
3.37k
| context
stringlengths 0
14.7k
|
|---|---|---|---|---|---|---|---|
pywhat/__dir__
|
Modified
|
bee-san~pyWhat
|
2f4e0ccb1d6b5508dc6ba61e40fb3dfe3b413ba0
|
Merge branch 'main' into flake8
|
<0>:<add> return __all__ + ["__version__"]
<del> return _contents + ["__version__"]
|
# module: pywhat
def __dir__():
<0> return _contents + ["__version__"]
<1>
|
===========changed ref 0===========
# module: pywhat
__version__ = "4.3.1"
tags = AvailableTags().get_tags()
pywhat_tags = tags # left for backward compatibility purposes
+ __all__ = ["Identifier", "Distribution", "tags", "pywhat_tags", "Keys", "Filter"]
- _contents = ["Identifier", "Distribution", "tags", "pywhat_tags", "Keys", "Filter"]
-
- __all__ = _contents
del AvailableTags, filter
===========changed ref 1===========
# module: pywhat.helper
class Keys(Enum):
+ def MATCHED(match):
+ return match["Matched"]
+
===========changed ref 2===========
# module: pywhat.helper
class Keys(Enum):
+ def NAME(match):
+ return match["Regex Pattern"]["Name"]
+
===========changed ref 3===========
# module: pywhat.helper
class Keys(Enum):
+ def RARITY(match):
+ return match["Regex Pattern"]["Rarity"]
+
===========changed ref 4===========
# module: pywhat.helper
class Keys(Enum):
- NAME = lambda match: match["Regex Pattern"]["Name"]
- RARITY = lambda match: match["Regex Pattern"]["Rarity"]
- MATCHED = lambda match: match["Matched"]
NONE = auto()
===========changed ref 5===========
# module: pywhat.filter
class Filter(Mapping):
def __contains__(self, item):
try:
return (
self["MinRarity"] <= item["Rarity"] <= self["MaxRarity"]
and set(item["Tags"]) & self["Tags"]
and not set(item["Tags"]) & self["ExcludeTags"]
)
+ except KeyError:
- except:
return False
|
pywhat.what/main
|
Modified
|
bee-san~pyWhat
|
2f4e0ccb1d6b5508dc6ba61e40fb3dfe3b413ba0
|
Merge branch 'main' into flake8
|
<s> pywhat.",
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
<0> """
<1> pyWhat - Identify what something is.
<2>
<3> Made by Bee https://twitter.com/bee_sec_san
<4>
<5> https://github.com/bee-san
<6>
<7> Filtration:
<8>
<9> --rarity min:max
<10>
<11> Rarity is how unlikely something is to be a false-positive. The higher the number, the more unlikely.
<12>
<13> Only print entries with rarity in range [min,max]. min and max can be omitted.
<14>
<15> Note: PyWhat by default has a rarity of 0.1. To see all matches, with many potential false positives use `0:`.
<16>
<17> --include list
<18>
<19> Only include entries containing at least one tag in a list. List is a comma separated list.
<20>
<21> --exclude list
<22>
<23> Exclude specified tags. List is a comma separated list.
<24>
<25> Sorting:
<26>
<27> --key key_name
<28>
<29> Sort by the given key.
<30>
<31> --reverse
<32>
<33> Sort in reverse order.
<34>
<35> Available keys:
<36>
<37> name - Sort by the name of regex pattern
<38>
<39> rarity - Sort by rarity
<40>
<41> matched - Sort by a matched string
<42>
<43> none - No sorting is done (the default)
<44>
<45> Exporting:
<46>
</s>
|
===========below chunk 0===========
<s>
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 1
Return results in json format.
Boundaryless mode:
CLI tool matches strings like 'abcdTHM{hello}plze' by default because the boundaryless mode is enabled for regexes with a rarity of 0.1 and higher.
Since boundaryless mode may produce a lot of false-positive matches, it is possible to disable it, either fully or partially.
'--disable-boundaryless' flag can be used to fully disable this mode.
In addition, '-br', '-bi', and '-be' options can be used to tweak which regexes should be in boundaryless mode.
Refer to the Filtration section for more information.
Formatting the output:
--format format_str
format_str can be equal to:
pretty - Output data in the table
json - Ouput data in json format
CUSTOM_STRING - Print data in the way you want. For every match CUSTOM_STRING will be printed and '%x' (See below for possible x values) will be substituted with a match value.
For example:
pywhat --format '%m - %n' 'google.com htb{flag}'
will print:
htb{flag} - HackTheBox Flag Format
google.com - Uniform Resource Locator (URL)
Possible '%x' values:
%m - matched text
%n - name of regex
%d - description</s>
===========below chunk 1===========
<s>
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 2
<s>
%m - matched text
%n - name of regex
%d - description (will not output if absent)
%e - exploit (will not ouput if absent)
%r - rarity
%l - link (will not ouput if absent)
%t - tags (in 'tag1, tag2 ...' format)
If you want to print '%' or '\' character - escape it: '\%', '\\'.
Examples:
* what 'HTB{this is a flag}'
* what '0x52908400098527886E0F7030069857D2E4169EE7'
* what -- '52.6169586, -1.9779857'
* what --rarity 0.6: '[email protected]'
* what --rarity 0: --include "credentials, username, password" --exclude "aws, credentials" 'James:SecretPassword'
* what -br 0.6: -be URL '[email protected]'
Your text must either be in quotation marks, or use the POSIX standard of "--" to mean "anything after -- is textual input".
pyWhat can also search files or even a whole directory with recursion</s>
===========below chunk 2===========
<s>
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 3
<s>
* what 'secret.txt'
* what 'this/is/a/path'
"""
if kwargs["text_input"] is None:
sys.exit("Text input expected. Run 'pywhat --help' for help")
dist = Distribution(
create_filter(kwargs["rarity"], kwargs["include"], kwargs["exclude"])
)
if kwargs["disable_boundaryless"]:
boundaryless = Filter({"Tags": []}) # use empty filter
else:
boundaryless = create_filter(
kwargs["boundaryless_rarity"],
kwargs["boundaryless_include"],
kwargs["boundaryless_exclude"],
)
what_obj = What_Object(dist)
if kwargs["key"] is None:
key = Keys.NONE
else:
try:
key = str_to_key(kwargs["key"])
except ValueError:
print("Invalid key")
sys.exit(1)
identified_output = what_obj.what_is_this(
kwargs["text_input"],
kwargs["only_text"],
key,
kwargs["reverse"],
boundaryless,
kwargs["include_filenames"],
)
p = printer.Printing()
</s>
===========below chunk 3===========
<s>
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 4
<s> if kwargs["json"] or str(kwargs["format"]).strip() == "json":
p.print_json(identified_output)
elif str(kwargs["format"]).strip() == "pretty":
p.pretty_print(identified_output, kwargs["text_input"], kwargs["print_tags"])
elif kwargs["format"] is not None:
p.format_print(identified_output, kwargs["format"])
else:
p.print_raw(identified_output, kwargs["text_input"], kwargs["print_tags"])
|
|
tests.test_filtration/test_distribution7
|
Modified
|
bee-san~pyWhat
|
2f4e0ccb1d6b5508dc6ba61e40fb3dfe3b413ba0
|
Merge branch 'main' into flake8
|
<1>:<add> Distribution({"Tags": "Media", "MinRarity": 0.7})
<del> dist = Distribution({"Tags": "Media", "MinRarity": 0.7})
|
# module: tests.test_filtration
def test_distribution7():
<0> with pytest.raises(InvalidTag):
<1> dist = Distribution({"Tags": "Media", "MinRarity": 0.7})
<2>
|
===========unchanged ref 0===========
at: _pytest.python_api
raises(expected_exception: Union[Type[E], Tuple[Type[E], ...]], func: Callable[..., Any], *args: Any, **kwargs: Any) -> _pytest._code.ExceptionInfo[E]
raises(expected_exception: Union[Type[E], Tuple[Type[E], ...]], *, match: Optional[Union[str, Pattern[str]]]=...) -> "RaisesContext[E]"
at: pywhat.filter
Distribution(filter: Optional[Filter]=None)
at: pywhat.helper
InvalidTag(*args: object)
===========changed ref 0===========
# module: pywhat.helper
class Keys(Enum):
+ def MATCHED(match):
+ return match["Matched"]
+
===========changed ref 1===========
# module: pywhat.helper
class Keys(Enum):
+ def NAME(match):
+ return match["Regex Pattern"]["Name"]
+
===========changed ref 2===========
# module: pywhat.helper
class Keys(Enum):
+ def RARITY(match):
+ return match["Regex Pattern"]["Rarity"]
+
===========changed ref 3===========
# module: pywhat
def __dir__():
+ return __all__ + ["__version__"]
- return _contents + ["__version__"]
===========changed ref 4===========
# module: pywhat.helper
class Keys(Enum):
- NAME = lambda match: match["Regex Pattern"]["Name"]
- RARITY = lambda match: match["Regex Pattern"]["Rarity"]
- MATCHED = lambda match: match["Matched"]
NONE = auto()
===========changed ref 5===========
# module: pywhat.filter
class Filter(Mapping):
def __contains__(self, item):
try:
return (
self["MinRarity"] <= item["Rarity"] <= self["MaxRarity"]
and set(item["Tags"]) & self["Tags"]
and not set(item["Tags"]) & self["ExcludeTags"]
)
+ except KeyError:
- except:
return False
===========changed ref 6===========
# module: pywhat
__version__ = "4.3.1"
tags = AvailableTags().get_tags()
pywhat_tags = tags # left for backward compatibility purposes
+ __all__ = ["Identifier", "Distribution", "tags", "pywhat_tags", "Keys", "Filter"]
- _contents = ["Identifier", "Distribution", "tags", "pywhat_tags", "Keys", "Filter"]
-
- __all__ = _contents
del AvailableTags, filter
===========changed ref 7===========
<s> pywhat.",
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
"""
pyWhat - Identify what something is.
Made by Bee https://twitter.com/bee_sec_san
https://github.com/bee-san
Filtration:
--rarity min:max
Rarity is how unlikely something is to be a false-positive. The higher the number, the more unlikely.
Only print entries with rarity in range [min,max]. min and max can be omitted.
Note: PyWhat by default has a rarity of 0.1. To see all matches, with many potential false positives use `0:`.
--include list
Only include entries containing at least one tag in a list. List is a comma separated list.
--exclude list
Exclude specified tags. List is a comma separated list.
Sorting:
--key key_name
Sort by the given key.
--reverse
Sort in reverse order.
Available keys:
name - Sort by the name of regex pattern
rarity - Sort by rarity
matched - Sort by a matched string
none - No sorting is done (the default)
Exporting:
--json
Return results in json format.
Boundaryless mode:
CLI tool matches strings like 'abcdTHM{hello}plze' by default because</s>
===========changed ref 8===========
<s>
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 1
<s>
Boundaryless mode:
CLI tool matches strings like 'abcdTHM{hello}plze' by default because the boundaryless mode is enabled for regexes with a rarity of 0.1 and higher.
Since boundaryless mode may produce a lot of false-positive matches, it is possible to disable it, either fully or partially.
'--disable-boundaryless' flag can be used to fully disable this mode.
In addition, '-br', '-bi', and '-be' options can be used to tweak which regexes should be in boundaryless mode.
Refer to the Filtration section for more information.
Formatting the output:
--format format_str
format_str can be equal to:
pretty - Output data in the table
json - Ouput data in json format
CUSTOM_STRING - Print data in the way you want. For every match CUSTOM_STRING will be printed and '%x' (See below for possible x values) will be substituted with a match value.
For example:
pywhat --format '%m - %n' 'google.com htb{flag}'
will print:
htb{flag} - HackTheBox Flag Format
google.com - Uniform Resource Locator (URL)
Possible '%x' values:
%m - matched</s>
|
tests.test_identifier/test_only_text
|
Modified
|
bee-san~pyWhat
|
2f4e0ccb1d6b5508dc6ba61e40fb3dfe3b413ba0
|
Merge branch 'main' into flake8
|
<1>:<add> assert out["Regexes"] is None
<del> assert None == out["Regexes"]
|
# module: tests.test_identifier
def test_only_text():
<0> out = r.identify("fixtures/file")
<1> assert None == out["Regexes"]
<2>
<3> out = r.identify("THM{7281j}}", only_text=True)
<4> assert "TryHackMe Flag Format" in out["Regexes"]["text"][0]["Regex Pattern"]["Name"]
<5>
|
===========unchanged ref 0===========
at: pywhat.identifier.Identifier
identify(text: str, *, only_text=True, dist: Distribution=None, key: Optional[Callable]=None, reverse: Optional[bool]=None, boundaryless: Optional[Filter]=None, include_filenames=False) -> dict
at: tests.test_identifier
r = identifier.Identifier()
===========changed ref 0===========
# module: pywhat.helper
class Keys(Enum):
+ def MATCHED(match):
+ return match["Matched"]
+
===========changed ref 1===========
# module: pywhat.helper
class Keys(Enum):
+ def NAME(match):
+ return match["Regex Pattern"]["Name"]
+
===========changed ref 2===========
# module: pywhat.helper
class Keys(Enum):
+ def RARITY(match):
+ return match["Regex Pattern"]["Rarity"]
+
===========changed ref 3===========
# module: tests.test_filtration
+ def test_invalid_contains_returns_false():
+ filter = Filter()
+ assert ({} in filter) is False
+
===========changed ref 4===========
# module: pywhat
def __dir__():
+ return __all__ + ["__version__"]
- return _contents + ["__version__"]
===========changed ref 5===========
# module: tests.test_filtration
def test_distribution7():
with pytest.raises(InvalidTag):
+ Distribution({"Tags": "Media", "MinRarity": 0.7})
- dist = Distribution({"Tags": "Media", "MinRarity": 0.7})
===========changed ref 6===========
# module: pywhat.helper
class Keys(Enum):
- NAME = lambda match: match["Regex Pattern"]["Name"]
- RARITY = lambda match: match["Regex Pattern"]["Rarity"]
- MATCHED = lambda match: match["Matched"]
NONE = auto()
===========changed ref 7===========
# module: pywhat.filter
class Filter(Mapping):
def __contains__(self, item):
try:
return (
self["MinRarity"] <= item["Rarity"] <= self["MaxRarity"]
and set(item["Tags"]) & self["Tags"]
and not set(item["Tags"]) & self["ExcludeTags"]
)
+ except KeyError:
- except:
return False
===========changed ref 8===========
# module: pywhat
__version__ = "4.3.1"
tags = AvailableTags().get_tags()
pywhat_tags = tags # left for backward compatibility purposes
+ __all__ = ["Identifier", "Distribution", "tags", "pywhat_tags", "Keys", "Filter"]
- _contents = ["Identifier", "Distribution", "tags", "pywhat_tags", "Keys", "Filter"]
-
- __all__ = _contents
del AvailableTags, filter
===========changed ref 9===========
<s> pywhat.",
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
"""
pyWhat - Identify what something is.
Made by Bee https://twitter.com/bee_sec_san
https://github.com/bee-san
Filtration:
--rarity min:max
Rarity is how unlikely something is to be a false-positive. The higher the number, the more unlikely.
Only print entries with rarity in range [min,max]. min and max can be omitted.
Note: PyWhat by default has a rarity of 0.1. To see all matches, with many potential false positives use `0:`.
--include list
Only include entries containing at least one tag in a list. List is a comma separated list.
--exclude list
Exclude specified tags. List is a comma separated list.
Sorting:
--key key_name
Sort by the given key.
--reverse
Sort in reverse order.
Available keys:
name - Sort by the name of regex pattern
rarity - Sort by rarity
matched - Sort by a matched string
none - No sorting is done (the default)
Exporting:
--json
Return results in json format.
Boundaryless mode:
CLI tool matches strings like 'abcdTHM{hello}plze' by default because</s>
===========changed ref 10===========
<s>
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 1
<s>
Boundaryless mode:
CLI tool matches strings like 'abcdTHM{hello}plze' by default because the boundaryless mode is enabled for regexes with a rarity of 0.1 and higher.
Since boundaryless mode may produce a lot of false-positive matches, it is possible to disable it, either fully or partially.
'--disable-boundaryless' flag can be used to fully disable this mode.
In addition, '-br', '-bi', and '-be' options can be used to tweak which regexes should be in boundaryless mode.
Refer to the Filtration section for more information.
Formatting the output:
--format format_str
format_str can be equal to:
pretty - Output data in the table
json - Ouput data in json format
CUSTOM_STRING - Print data in the way you want. For every match CUSTOM_STRING will be printed and '%x' (See below for possible x values) will be substituted with a match value.
For example:
pywhat --format '%m - %n' 'google.com htb{flag}'
will print:
htb{flag} - HackTheBox Flag Format
google.com - Uniform Resource Locator (URL)
Possible '%x' values:
%m - matched</s>
|
pywhat.filter/Filter.__contains__
|
Modified
|
bee-san~pyWhat
|
4342bdc6011afa30dd03a196c7804ead223d3984
|
refactor: remove unnecessary try except
|
<0>:<del> try:
<1>:<add> return (
<del> return (
<2>:<add> self["MinRarity"] <= item["Rarity"] <= self["MaxRarity"]
<del> self["MinRarity"] <= item["Rarity"] <= self["MaxRarity"]
<3>:<add> and set(item["Tags"]) & self["Tags"]
<del> and set(item["Tags"]) & self["Tags"]
<4>:<add> and not set(item["Tags"]) & self["ExcludeTags"]
<del> and not set(item["Tags"]) & self["ExcludeTags"]
<5>:<add> )
<del> )
<6>:<del> except KeyError:
<7>:<del> return False
|
# module: pywhat.filter
class Filter(Mapping):
def __contains__(self, item):
<0> try:
<1> return (
<2> self["MinRarity"] <= item["Rarity"] <= self["MaxRarity"]
<3> and set(item["Tags"]) & self["Tags"]
<4> and not set(item["Tags"]) & self["ExcludeTags"]
<5> )
<6> except KeyError:
<7> return False
<8>
|
===========unchanged ref 0===========
at: pywhat.filter.Filter.__init__
self._dict = {}
at: typing.Mapping
__contains__(self, o: object) -> bool
|
pywhat.regex_identifier/RegexIdentifier.check
|
Modified
|
bee-san~pyWhat
|
a80cabdcee42a8dc09ef4a5098b61b063dd58473
|
Merge branch 'main' into perf-regex-identifier-check-shallow
|
<12>:<add> reg_match = dict(reg)
<del> reg_match = copy.deepcopy(reg)
|
# module: pywhat.regex_identifier
class RegexIdentifier:
def check(
self,
text,
dist: Optional[Distribution] = None,
*,
boundaryless: Optional[Filter] = None
):
<0> if dist is None:
<1> dist = self.distribution
<2> if boundaryless is None:
<3> boundaryless = Filter({"Tags": []})
<4> matches = []
<5>
<6> for string in text:
<7> for reg in dist.get_regexes():
<8> regex = (
<9> reg["Boundaryless Regex"] if reg in boundaryless else reg["Regex"]
<10> )
<11> for matched_regex in re.finditer(regex, string, re.MULTILINE):
<12> reg_match = copy.deepcopy(reg)
<13> matched = self.clean_text(matched_regex.group(0))
<14>
<15> if (
<16> reg_match.get("Exploit") is not None
<17> and "curl" in reg_match["Exploit"]
<18> ):
<19> # Replace anything like XXXXX_XXXXXX_HERE with the match
<20> reg_match["Exploit"] = re.sub(
<21> r"[A-Z_]+_HERE", matched, reg_match["Exploit"]
<22> )
<23>
<24> children = reg_match.get("Children")
<25> if children is not None:
<26> processed_match = re.sub(
<27> children.get("deletion_pattern", ""), "", matched
<28> )
<29> matched_children = []
<30> if children["method"] == "hashmap":
<31> for length in children["lengths"]:
<32> try:
<33> matched_children.append(
<34> children["Items"][processed_match[:length]]
<35> )
<36> except KeyError:
<37> continue
<38> else:
<39> for element in children["Items"]:
<40> if (
<41> children["method"] == "regex"
<42> and re.search(
<43> element, processed_match, re.MULTILINE
<44> )
<45> )</s>
|
===========below chunk 0===========
# module: pywhat.regex_identifier
class RegexIdentifier:
def check(
self,
text,
dist: Optional[Distribution] = None,
*,
boundaryless: Optional[Filter] = None
):
# offset: 1
children["method"] == "startswith"
and processed_match.startswith(element)
):
matched_children.append(children["Items"][element])
if matched_children:
reg_match["Description"] = children.get(
"entry", ""
) + ", ".join(matched_children)
reg_match.pop("Children", None)
matches.append(
{
"Matched": matched,
"Regex Pattern": reg_match,
}
)
return matches
===========unchanged ref 0===========
at: pywhat.filter
Filter(filters_dict=None)
Distribution(filter: Optional[Filter]=None)
at: pywhat.filter.Distribution
get_regexes()
at: pywhat.regex_identifier.RegexIdentifier
clean_text(text)
at: pywhat.regex_identifier.RegexIdentifier.__init__
self.distribution = Distribution()
at: re
MULTILINE = RegexFlag.MULTILINE
search(pattern: Pattern[AnyStr], string: AnyStr, flags: _FlagsType=...) -> Optional[Match[AnyStr]]
search(pattern: AnyStr, string: AnyStr, flags: _FlagsType=...) -> Optional[Match[AnyStr]]
sub(pattern: AnyStr, repl: AnyStr, string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: Pattern[AnyStr], repl: Callable[[Match[AnyStr]], AnyStr], string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: AnyStr, repl: Callable[[Match[AnyStr]], AnyStr], string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: Pattern[AnyStr], repl: AnyStr, string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
finditer(pattern: AnyStr, string: AnyStr, flags: _FlagsType=...) -> Iterator[Match[AnyStr]]
finditer(pattern: Pattern[AnyStr], string: AnyStr, flags: _FlagsType=...) -> Iterator[Match[AnyStr]]
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
at: typing.Match
pos: int
endpos: int
lastindex: Optional[int]
lastgroup: Optional[AnyStr]
string: AnyStr
re: Pattern[AnyStr]
===========unchanged ref 1===========
group(group1: Union[str, int], group2: Union[str, int], /, *groups: Union[str, int]) -> Tuple[AnyStr, ...]
group(group: Union[str, int]=..., /) -> AnyStr
|
noxfile/tests
|
Modified
|
bee-san~pyWhat
|
67de66b1b5e4c0ad1d56cb1143f622c3e54e98f8
|
Merge branch 'main' into main
|
<8>:<add> "pytest-flake8",
|
# module: noxfile
@nox.session
def tests(session: Session) -> None:
<0> """Run the test suite."""
<1> session.run("poetry", "install", "--no-dev", external=True)
<2> install_with_constraints(
<3> session,
<4> "pytest",
<5> "pytest-black",
<6> "pytest-cov",
<7> "pytest-isort",
<8> "pytest-mypy",
<9> "types-requests",
<10> "types-orjson",
<11> )
<12> session.run("pytest", "--cov=./", "--cov-report=xml")
<13>
|
===========unchanged ref 0===========
at: noxfile
install_with_constraints(session: Session, *args: str, **kwargs: Any) -> None
|
pywhat.what/main
|
Modified
|
bee-san~pyWhat
|
9ae9fcb770af3eb5769ca28510e92be52465b8e2
|
Merge branch 'main' into fix-ec2
|
<s> is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
+ @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
- @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
<0> """
<1> pyWhat - Identify what something is.
<2>
<3> Made by Bee https://twitter.com/bee_sec_san
<4>
<5> https://github.com/bee-san
<6>
<7> Filtration:
<8>
<9> --rarity min:max
<10>
<11> Rarity is how unlikely something is to be a false-positive. The higher the number, the more unlikely.
<12>
<13> Only print entries with rarity in range [min,max]. min and max can be omitted.
<14>
<15> Note: PyWhat by default has a rarity of 0.1. To see all matches, with many potential false positives use `0:`.
<16>
<17> --include list
<18>
<19> Only include entries containing at least one tag in a list. List is a comma separated list.
<20>
<21> --exclude list
<22>
<23> Exclude specified tags. List is a comma separated list.
<24>
<25> Sorting:
<26>
<27> --key key_name
<28>
<29> Sort by the given key.
<30>
<31> --reverse
<32>
<33> Sort in reverse order.
<34>
<35> Available keys:
<36>
<37> name - Sort by the name of regex pattern
<38>
<39> rarity - Sort by rarity
<40>
<41> matched - Sort by a matched string
<42>
<43> none - No sorting is done (the default)
<44>
<45> Exporting:
<46>
</s>
|
===========below chunk 0===========
<s>True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
+ @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
- @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 1
Return results in json format.
Boundaryless mode:
CLI tool matches strings like 'abcdTHM{hello}plze' by default because the boundaryless mode is enabled for regexes with a rarity of 0.1 and higher.
Since boundaryless mode may produce a lot of false-positive matches, it is possible to disable it, either fully or partially.
'--disable-boundaryless' flag can be used to fully disable this mode.
In addition, '-br', '-bi', and '-be' options can be used to tweak which regexes should be in boundaryless mode.
Refer to the Filtration section for more information.
Formatting the output:
--format format_str
format_str can be equal to:
pretty - Output data in the table
json - Ouput data in json format
CUSTOM_STRING - Print data in the way you want. For every match CUSTOM_STRING will be printed and '%x' (See below for possible x values) will be substituted with a match value.
For example:
pywhat --format '%m - %n' 'google.com htb{flag}'
will print:
htb{flag} - HackTheBox Flag Format
google.com - Uniform Resource Locator (URL)
Possible '%x' values:
%m - matched text
%n - name of regex
%d - description</s>
===========below chunk 1===========
<s>True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
+ @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
- @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 2
<s>
%m - matched text
%n - name of regex
%d - description (will not output if absent)
%e - exploit (will not ouput if absent)
%r - rarity
%l - link (will not ouput if absent)
%t - tags (in 'tag1, tag2 ...' format)
If you want to print '%' or '\\' character - escape it: '\\%', '\\\\'.
Examples:
* what 'HTB{this is a flag}'
* what '0x52908400098527886E0F7030069857D2E4169EE7'
* what -- '52.6169586, -1.9779857'
* what --rarity 0.6: '[email protected]'
* what --rarity 0: --include "credentials, username, password" --exclude "aws, credentials" 'James:SecretPassword'
* what -br 0.6: -be URL '[email protected]'
Your text must either be in quotation marks, or use the POSIX standard of "--" to mean "anything after -- is textual input".
pyWhat can also search files or even a whole directory with</s>
===========below chunk 2===========
<s>True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
+ @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
- @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 3
<s>
* what 'secret.txt'
* what 'this/is/a/path'
"""
if kwargs["text_input"] is None:
sys.exit("Text input expected. Run 'pywhat --help' for help")
dist = Distribution(
create_filter(kwargs["rarity"], kwargs["include"], kwargs["exclude"])
)
if kwargs["disable_boundaryless"]:
boundaryless = Filter({"Tags": []}) # use empty filter
else:
boundaryless = create_filter(
kwargs["boundaryless_rarity"],
kwargs["boundaryless_include"],
kwargs["boundaryless_exclude"],
)
what_obj = What_Object(dist)
if kwargs["key"] is None:
key = Keys.NONE
else:
try:
key = str_to_key(kwargs["key"])
except ValueError:
print("Invalid key")
sys.exit(1)
identified_output = what_obj.what_is_this(
kwargs["text_input"],
kwargs["only_text"],
key,
kwargs["reverse"],
boundaryless,
kwargs["include_filenames"],
)
p = printer.Printing()
</s>
===========below chunk 3===========
<s>True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
+ @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
- @click.option("-pt", "--print-tags", is_flag=True, help="Add flags to ouput")
def main(**kwargs):
# offset: 4
<s> if kwargs["json"] or str(kwargs["format"]).strip() == "json":
p.print_json(identified_output)
elif str(kwargs["format"]).strip() == "pretty":
p.pretty_print(identified_output, kwargs["text_input"], kwargs["print_tags"])
elif kwargs["format"] is not None:
p.format_print(identified_output, kwargs["format"])
else:
p.print_raw(identified_output, kwargs["text_input"], kwargs["print_tags"])
|
|
pywhat.printer/Printing.print_json
|
Modified
|
bee-san~pyWhat
|
c89c57ca973bc4b94bdf1290067fa98db49fe61a
|
Fix json printing modified regex
|
<0>:<add> print(json.dumps(text))
<del> self.console.print(json.dumps(text))
|
# module: pywhat.printer
class Printing:
def print_json(self, text: dict):
<0> self.console.print(json.dumps(text))
<1>
|
===========unchanged ref 0===========
at: json
dumps(obj: Any, *, skipkeys: bool=..., ensure_ascii: bool=..., check_circular: bool=..., allow_nan: bool=..., cls: Optional[Type[JSONEncoder]]=..., indent: Union[None, int, str]=..., separators: Optional[Tuple[str, str]]=..., default: Optional[Callable[[Any], Any]]=..., sort_keys: bool=..., **kwds: Any) -> str
|
pywhat.what/main
|
Modified
|
bee-san~pyWhat
|
e2e081b0cebf9785eb0e38efb1b1245de151d51f
|
Fix example invocation with tags in help string
|
<s> version of pywhat.",
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
def main(**kwargs):
<0> """
<1> pyWhat - Identify what something is.
<2>
<3> Made by Bee https://twitter.com/bee_sec_san
<4>
<5> https://github.com/bee-san
<6>
<7> Filtration:
<8>
<9> --rarity min:max
<10>
<11> Rarity is how unlikely something is to be a false-positive. The higher the number, the more unlikely.
<12>
<13> Only print entries with rarity in range [min,max]. min and max can be omitted.
<14>
<15> Note: PyWhat by default has a rarity of 0.1. To see all matches, with many potential false positives use `0:`.
<16>
<17> --include list
<18>
<19> Only include entries containing at least one tag in a list. List is a comma separated list.
<20>
<21> --exclude list
<22>
<23> Exclude specified tags. List is a comma separated list.
<24>
<25> Sorting:
<26>
<27> --key key_name
<28>
<29> Sort by the given key.
<30>
<31> --reverse
<32>
<33> Sort in reverse order.
<34>
<35> Available keys:
<36>
<37> name - Sort by the name of regex pattern
<38>
<39> rarity - Sort by rarity
<40>
<41> matched - Sort by a matched string
<42>
<43> none - No sorting is done (the default)
<44>
<45> Exporting:
<46>
</s>
|
===========below chunk 0===========
<s>,
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
def main(**kwargs):
# offset: 1
Return results in json format.
Boundaryless mode:
CLI tool matches strings like 'abcdTHM{hello}plze' by default because the boundaryless mode is enabled for regexes with a rarity of 0.1 and higher.
Since boundaryless mode may produce a lot of false-positive matches, it is possible to disable it, either fully or partially.
'--disable-boundaryless' flag can be used to fully disable this mode.
In addition, '-br', '-bi', and '-be' options can be used to tweak which regexes should be in boundaryless mode.
Refer to the Filtration section for more information.
Formatting the output:
--format format_str
format_str can be equal to:
pretty - Output data in the table
json - Output data in json format
CUSTOM_STRING - Print data in the way you want. For every match CUSTOM_STRING will be printed and '%x' (See below for possible x values) will be substituted with a match value.
For example:
pywhat --format '%m - %n' 'google.com htb{flag}'
will print:
htb{flag} - HackTheBox Flag Format
google.com - Uniform Resource Locator (URL)
Possible '%x' values:
%m - matched text
%n - name of regex
%d - description (will</s>
===========below chunk 1===========
<s>,
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
def main(**kwargs):
# offset: 2
<s>
%m - matched text
%n - name of regex
%d - description (will not output if absent)
%e - exploit (will not output if absent)
%r - rarity
%l - link (will not output if absent)
%t - tags (in 'tag1, tag2 ...' format)
If you want to print '%' or '\\' character - escape it: '\\%', '\\\\'.
Examples:
* what 'HTB{this is a flag}'
* what '0x52908400098527886E0F7030069857D2E4169EE7'
* what -- '52.6169586, -1.9779857'
* what --rarity 0.6: '[email protected]'
* what --rarity 0: --include "credentials, username, password" --exclude "aws, credentials" 'James:SecretPassword'
* what -br 0.6: -be URL '[email protected]'
Your text must either be in quotation marks, or use the POSIX standard of "--" to mean "anything after -- is textual input".
pyWhat can also search files or even a whole directory with recursion:
</s>
===========below chunk 2===========
<s>,
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
def main(**kwargs):
# offset: 3
<s> what 'secret.txt'
* what 'this/is/a/path'
"""
if kwargs["text_input"] is None:
sys.exit("Text input expected. Run 'pywhat --help' for help")
dist = Distribution(
create_filter(kwargs["rarity"], kwargs["include"], kwargs["exclude"])
)
if kwargs["disable_boundaryless"]:
boundaryless = Filter({"Tags": []}) # use empty filter
else:
boundaryless = create_filter(
kwargs["boundaryless_rarity"],
kwargs["boundaryless_include"],
kwargs["boundaryless_exclude"],
)
what_obj = What_Object(dist)
if kwargs["key"] is None:
key = Keys.NONE
else:
try:
key = str_to_key(kwargs["key"])
except ValueError:
print("Invalid key")
sys.exit(1)
identified_output = what_obj.what_is_this(
kwargs["text_input"],
kwargs["only_text"],
key,
kwargs["reverse"],
boundaryless,
kwargs["include_filenames"],
)
p = printer.Printing()
if kwargs["</s>
===========below chunk 3===========
<s>,
)
@click.option(
"-if",
"--include-filenames",
is_flag=True,
help="Search filenames for possible matches.",
)
@click.option(
"--format",
required=False,
help="Format output according to specified rules.",
)
@click.option("-pt", "--print-tags", is_flag=True, help="Add flags to output")
def main(**kwargs):
# offset: 4
<s> or str(kwargs["format"]).strip() == "json":
p.print_json(identified_output)
elif str(kwargs["format"]).strip() == "pretty":
p.pretty_print(identified_output, kwargs["text_input"], kwargs["print_tags"])
elif kwargs["format"] is not None:
p.format_print(identified_output, kwargs["format"])
else:
p.print_raw(identified_output, kwargs["text_input"], kwargs["print_tags"])
|
|
scripts.prepdocs/create_sections
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
e201432ac6878411518b5c16c21bddcf39bd42ae
|
Remove unsupported characters from Azure Search section id (#39)
|
<2>:<add> "id": re.sub("[^0-9a-zA-Z_-]","_",f"{filename}-{i}"),
<del> "id": f"{filename}-{i}".replace(".", "_").replace(" ", "_"),
|
# module: scripts.prepdocs
def create_sections(filename, pages):
<0> for i, (section, pagenum) in enumerate(split_text(pages)):
<1> yield {
<2> "id": f"{filename}-{i}".replace(".", "_").replace(" ", "_"),
<3> "content": section,
<4> "category": args.category,
<5> "sourcepage": blob_name_from_file_page(filename, pagenum),
<6> "sourcefile": filename
<7> }
<8>
|
===========unchanged ref 0===========
at: re
sub(pattern: AnyStr, repl: AnyStr, string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: Pattern[AnyStr], repl: Callable[[Match[AnyStr]], AnyStr], string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: AnyStr, repl: Callable[[Match[AnyStr]], AnyStr], string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: Pattern[AnyStr], repl: AnyStr, string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
at: scripts.prepdocs
args = parser.parse_args()
blob_name_from_file_page(filename, page)
split_text(pages)
|
scripts.prepdocs/index_sections
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
1273a2111e92263d53a9ffa8d6abfd0531961c9b
|
Fix typo which causes error for document with more than 1000 sections. (#38)
|
<10>:<add> results = search_client.upload_documents(documents=batch)
<del> results = search_client.index_documents(batch=batch)
|
# module: scripts.prepdocs
def index_sections(filename, sections):
<0> if args.verbose: print(f"Indexing sections from '{filename}' into search index '{args.index}'")
<1> search_client = SearchClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
<2> index_name=args.index,
<3> credential=search_creds)
<4> i = 0
<5> batch = []
<6> for s in sections:
<7> batch.append(s)
<8> i += 1
<9> if i % 1000 == 0:
<10> results = search_client.index_documents(batch=batch)
<11> succeeded = sum([1 for r in results if r.succeeded])
<12> if args.verbose: print(f"\tIndexed {len(results)} sections, {succeeded} succeeded")
<13> batch = []
<14>
<15> if len(batch) > 0:
<16> results = search_client.upload_documents(documents=batch)
<17> succeeded = sum([1 for r in results if r.succeeded])
<18> if args.verbose: print(f"\tIndexed {len(results)} sections, {succeeded} succeeded")
<19>
|
===========unchanged ref 0===========
at: scripts.prepdocs
args = parser.parse_args()
search_creds = default_creds if args.searchkey == None else AzureKeyCredential(args.searchkey)
|
scripts.prepdocs/blob_name_from_file_page
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6ac7c909c02d760bafd5e5e838fa8c2a46dd4aaf
|
Use Azure Form Recognizer as document preprocessing to extract text, tables, and document layout (#37)
|
<0>:<add> if os.path.splitext(filename)[1].lower() == ".pdf":
<add> return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
<del> return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
<1>:<add> else:
<add> return os.path.basename(filename)
|
# module: scripts.prepdocs
+ def blob_name_from_file_page(filename, page = 0):
- def blob_name_from_file_page(filename, page):
<0> return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
<1>
|
===========unchanged ref 0===========
at: scripts.prepdocs
args = parser.parse_args()
azd_credential = AzureDeveloperCliCredential() if args.tenantid == None else AzureDeveloperCliCredential(tenant_id=args.tenantid)
===========changed ref 0===========
# module: scripts.prepdocs
MAX_SECTION_LENGTH = 1000
SENTENCE_SEARCH_LIMIT = 100
SECTION_OVERLAP = 100
parser = argparse.ArgumentParser(
description="Prepare documents by extracting content from PDFs, splitting content into sections, uploading to blob storage, and indexing in a search index.",
epilog="Example: prepdocs.py '..\data\*' --storageaccount myaccount --container mycontainer --searchservice mysearch --index myindex -v"
)
parser.add_argument("files", help="Files to be processed")
parser.add_argument("--category", help="Value for the category field in the search index for all sections indexed in this run")
parser.add_argument("--skipblobs", action="store_true", help="Skip uploading individual pages to Azure Blob Storage")
parser.add_argument("--storageaccount", help="Azure Blob Storage account name")
parser.add_argument("--container", help="Azure Blob Storage container name")
parser.add_argument("--storagekey", required=False, help="Optional. Use this Azure Blob Storage account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--tenantid", required=False, help="Optional. Use this to define the Azure directory where to authenticate)")
parser.add_argument("--searchservice", help="Name of the Azure Cognitive Search service where content should be indexed (must exist already)")
parser.add_argument("--index", help="Name of the Azure Cognitive Search index where content should be indexed (will be created if it doesn't exist)")
parser.add_argument("--searchkey", required=False, help="Optional. Use this Azure Cognitive Search account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--remove", action="store_true", help="Remove references to this document from blob storage and the search index")
parser.add_argument("--removeall", action="store_true", help="Remove all blobs from blob storage and documents from the search index")
+ parser.add_argument("--localpdfparser", action="store_true", help="Use PyPdf local PDF parser (</s>
===========changed ref 1===========
# module: scripts.prepdocs
# offset: 1
<s> index")
+ parser.add_argument("--localpdfparser", action="store_true", help="Use PyPdf local PDF parser (supports only digital PDFs) instead of Azure Form Recognizer service to extract text, tables and layout from the documents")
+ parser.add_argument("--formrecognizerservice", required=False, help="Optional. Name of the Azure Form Recognizer service which will be used to extract text, tables and layout from the documents (must exist already)")
+ parser.add_argument("--formrecognizerkey", required=False, help="Optional. Use this Azure Form Recognizer account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--verbose", "-v", action="store_true", help="Verbose output")
args = parser.parse_args()
# Use the current user identity to connect to Azure services unless a key is explicitly set for any of them
azd_credential = AzureDeveloperCliCredential() if args.tenantid == None else AzureDeveloperCliCredential(tenant_id=args.tenantid)
default_creds = azd_credential if args.searchkey == None or args.storagekey == None else None
search_creds = default_creds if args.searchkey == None else AzureKeyCredential(args.searchkey)
if not args.skipblobs:
storage_creds = default_creds if args.storagekey == None else args.storagekey
+ if not args.localpdfparser:
+ # check if Azure Form Recognizer credentials are provided
+ if args.formrecognizerservice == None:
+ print("Error: Azure Form Recognizer service is not provided. Please provide formrecognizerservice or use --localpdfparser for local pypdf parser.")
+ exit(1)
+ formrecognizer_creds = default_creds if args.formrecognizerkey == None else AzureKeyCredential(args.formrecognizerkey)
|
scripts.prepdocs/upload_blobs
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6ac7c909c02d760bafd5e5e838fa8c2a46dd4aaf
|
Use Azure Form Recognizer as document preprocessing to extract text, tables, and document layout (#37)
|
<4>:<del> for i in range(len(pages)):
<5>:<del> blob_name = blob_name_from_file_page(filename, i)
<6>:<del> if args.verbose: print(f"\tUploading blob for page {i} -> {blob_name}")
<7>:<del> f = io.BytesIO()
<8>:<del> writer = PdfWriter()
<9>:<del> writer.add_page(pages[i])
<10>:<del> writer.write(f)
<11>:<del> f.seek(0)
<12>:<del> blob_container.upload_blob(blob_name, f, overwrite=True)
|
# module: scripts.prepdocs
+ def upload_blobs(filename):
- def upload_blobs(pages):
<0> blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
<1> blob_container = blob_service.get_container_client(args.container)
<2> if not blob_container.exists():
<3> blob_container.create_container()
<4> for i in range(len(pages)):
<5> blob_name = blob_name_from_file_page(filename, i)
<6> if args.verbose: print(f"\tUploading blob for page {i} -> {blob_name}")
<7> f = io.BytesIO()
<8> writer = PdfWriter()
<9> writer.add_page(pages[i])
<10> writer.write(f)
<11> f.seek(0)
<12> blob_container.upload_blob(blob_name, f, overwrite=True)
<13>
|
===========unchanged ref 0===========
at: os.path
splitext(p: AnyStr) -> Tuple[AnyStr, AnyStr]
splitext(p: _PathLike[AnyStr]) -> Tuple[AnyStr, AnyStr]
basename(p: _PathLike[AnyStr]) -> AnyStr
basename(p: AnyStr) -> AnyStr
at: scripts.prepdocs
args = parser.parse_args()
default_creds = azd_credential if args.searchkey == None or args.storagekey == None else None
===========changed ref 0===========
# module: scripts.prepdocs
+ def blob_name_from_file_page(filename, page = 0):
- def blob_name_from_file_page(filename, page):
+ if os.path.splitext(filename)[1].lower() == ".pdf":
+ return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
- return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
+ else:
+ return os.path.basename(filename)
===========changed ref 1===========
# module: scripts.prepdocs
MAX_SECTION_LENGTH = 1000
SENTENCE_SEARCH_LIMIT = 100
SECTION_OVERLAP = 100
parser = argparse.ArgumentParser(
description="Prepare documents by extracting content from PDFs, splitting content into sections, uploading to blob storage, and indexing in a search index.",
epilog="Example: prepdocs.py '..\data\*' --storageaccount myaccount --container mycontainer --searchservice mysearch --index myindex -v"
)
parser.add_argument("files", help="Files to be processed")
parser.add_argument("--category", help="Value for the category field in the search index for all sections indexed in this run")
parser.add_argument("--skipblobs", action="store_true", help="Skip uploading individual pages to Azure Blob Storage")
parser.add_argument("--storageaccount", help="Azure Blob Storage account name")
parser.add_argument("--container", help="Azure Blob Storage container name")
parser.add_argument("--storagekey", required=False, help="Optional. Use this Azure Blob Storage account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--tenantid", required=False, help="Optional. Use this to define the Azure directory where to authenticate)")
parser.add_argument("--searchservice", help="Name of the Azure Cognitive Search service where content should be indexed (must exist already)")
parser.add_argument("--index", help="Name of the Azure Cognitive Search index where content should be indexed (will be created if it doesn't exist)")
parser.add_argument("--searchkey", required=False, help="Optional. Use this Azure Cognitive Search account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--remove", action="store_true", help="Remove references to this document from blob storage and the search index")
parser.add_argument("--removeall", action="store_true", help="Remove all blobs from blob storage and documents from the search index")
+ parser.add_argument("--localpdfparser", action="store_true", help="Use PyPdf local PDF parser (</s>
===========changed ref 2===========
# module: scripts.prepdocs
# offset: 1
<s> index")
+ parser.add_argument("--localpdfparser", action="store_true", help="Use PyPdf local PDF parser (supports only digital PDFs) instead of Azure Form Recognizer service to extract text, tables and layout from the documents")
+ parser.add_argument("--formrecognizerservice", required=False, help="Optional. Name of the Azure Form Recognizer service which will be used to extract text, tables and layout from the documents (must exist already)")
+ parser.add_argument("--formrecognizerkey", required=False, help="Optional. Use this Azure Form Recognizer account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--verbose", "-v", action="store_true", help="Verbose output")
args = parser.parse_args()
# Use the current user identity to connect to Azure services unless a key is explicitly set for any of them
azd_credential = AzureDeveloperCliCredential() if args.tenantid == None else AzureDeveloperCliCredential(tenant_id=args.tenantid)
default_creds = azd_credential if args.searchkey == None or args.storagekey == None else None
search_creds = default_creds if args.searchkey == None else AzureKeyCredential(args.searchkey)
if not args.skipblobs:
storage_creds = default_creds if args.storagekey == None else args.storagekey
+ if not args.localpdfparser:
+ # check if Azure Form Recognizer credentials are provided
+ if args.formrecognizerservice == None:
+ print("Error: Azure Form Recognizer service is not provided. Please provide formrecognizerservice or use --localpdfparser for local pypdf parser.")
+ exit(1)
+ formrecognizer_creds = default_creds if args.formrecognizerkey == None else AzureKeyCredential(args.formrecognizerkey)
|
scripts.prepdocs/split_text
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6ac7c909c02d760bafd5e5e838fa8c2a46dd4aaf
|
Use Azure Form Recognizer as document preprocessing to extract text, tables, and document layout (#37)
|
<3>:<del>
<4>:<del> page_map = []
<5>:<del> offset = 0
<6>:<del> for i, p in enumerate(pages):
<7>:<del> text = p.extract_text()
<8>:<del> page_map.append((i, offset, text))
<9>:<del> offset += len(text)
|
# module: scripts.prepdocs
+ def split_text(page_map):
- def split_text(pages):
<0> SENTENCE_ENDINGS = [".", "!", "?"]
<1> WORDS_BREAKS = [",", ";", ":", " ", "(", ")", "[", "]", "{", "}", "\t", "\n"]
<2> if args.verbose: print(f"Splitting '{filename}' into sections")
<3>
<4> page_map = []
<5> offset = 0
<6> for i, p in enumerate(pages):
<7> text = p.extract_text()
<8> page_map.append((i, offset, text))
<9> offset += len(text)
<10>
<11> def find_page(offset):
<12> l = len(page_map)
<13> for i in range(l - 1):
<14> if offset >= page_map[i][1] and offset < page_map[i + 1][1]:
<15> return i
<16> return l - 1
<17>
<18> all_text = "".join(p[2] for p in page_map)
<19> length = len(all_text)
<20> start = 0
<21> end = length
<22> while start + SECTION_OVERLAP < length:
<23> last_word = -1
<24> end = start + MAX_SECTION_LENGTH
<25>
<26> if end > length:
<27> end = length
<28> else:
<29> # Try to find the end of the sentence
<30> while end < length and (end - start - MAX_SECTION_LENGTH) < SENTENCE_SEARCH_LIMIT and all_text[end] not in SENTENCE_ENDINGS:
<31> if all_text[end] in WORDS_BREAKS:
<32> last_word = end
<33> end += 1
<34> if end < length and all_text[end] not in SENTENCE_ENDINGS and last_word > 0:
<35> end = last_word # Fall back to at least keeping a whole word
<36> if end < length:
<37> end += 1
<38>
<39> # Try to find the start of the sentence or at least a whole word boundary
<40> </s>
|
===========below chunk 0===========
# module: scripts.prepdocs
+ def split_text(page_map):
- def split_text(pages):
# offset: 1
while start > 0 and start > end - MAX_SECTION_LENGTH - 2 * SENTENCE_SEARCH_LIMIT and all_text[start] not in SENTENCE_ENDINGS:
if all_text[start] in WORDS_BREAKS:
last_word = start
start -= 1
if all_text[start] not in SENTENCE_ENDINGS and last_word > 0:
start = last_word
if start > 0:
start += 1
yield (all_text[start:end], find_page(start))
start = end - SECTION_OVERLAP
if start + SECTION_OVERLAP < end:
yield (all_text[start:end], find_page(start))
===========unchanged ref 0===========
at: html
escape(s: AnyStr, quote: bool=...) -> AnyStr
at: io.BytesIO
seek(self, offset: int, whence: int=..., /) -> int
at: os.path
splitext(p: AnyStr) -> Tuple[AnyStr, AnyStr]
splitext(p: _PathLike[AnyStr]) -> Tuple[AnyStr, AnyStr]
basename(p: _PathLike[AnyStr]) -> AnyStr
basename(p: AnyStr) -> AnyStr
at: re
match(pattern: AnyStr, string: AnyStr, flags: _FlagsType=...) -> Optional[Match[AnyStr]]
match(pattern: Pattern[AnyStr], string: AnyStr, flags: _FlagsType=...) -> Optional[Match[AnyStr]]
at: scripts.prepdocs
args = parser.parse_args()
storage_creds = default_creds if args.storagekey == None else args.storagekey
formrecognizer_creds = default_creds if args.formrecognizerkey == None else AzureKeyCredential(args.formrecognizerkey)
blob_name_from_file_page(filename, page=0)
at: scripts.prepdocs.upload_blobs
blob_container = blob_service.get_container_client(args.container)
pages = reader.pages
f = io.BytesIO()
writer = PdfWriter()
===========changed ref 0===========
# module: scripts.prepdocs
+ def blob_name_from_file_page(filename, page = 0):
- def blob_name_from_file_page(filename, page):
+ if os.path.splitext(filename)[1].lower() == ".pdf":
+ return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
- return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
+ else:
+ return os.path.basename(filename)
===========changed ref 1===========
# module: scripts.prepdocs
+ def upload_blobs(filename):
- def upload_blobs(pages):
blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
blob_container = blob_service.get_container_client(args.container)
if not blob_container.exists():
blob_container.create_container()
- for i in range(len(pages)):
- blob_name = blob_name_from_file_page(filename, i)
- if args.verbose: print(f"\tUploading blob for page {i} -> {blob_name}")
- f = io.BytesIO()
- writer = PdfWriter()
- writer.add_page(pages[i])
- writer.write(f)
- f.seek(0)
- blob_container.upload_blob(blob_name, f, overwrite=True)
===========changed ref 2===========
# module: scripts.prepdocs
MAX_SECTION_LENGTH = 1000
SENTENCE_SEARCH_LIMIT = 100
SECTION_OVERLAP = 100
parser = argparse.ArgumentParser(
description="Prepare documents by extracting content from PDFs, splitting content into sections, uploading to blob storage, and indexing in a search index.",
epilog="Example: prepdocs.py '..\data\*' --storageaccount myaccount --container mycontainer --searchservice mysearch --index myindex -v"
)
parser.add_argument("files", help="Files to be processed")
parser.add_argument("--category", help="Value for the category field in the search index for all sections indexed in this run")
parser.add_argument("--skipblobs", action="store_true", help="Skip uploading individual pages to Azure Blob Storage")
parser.add_argument("--storageaccount", help="Azure Blob Storage account name")
parser.add_argument("--container", help="Azure Blob Storage container name")
parser.add_argument("--storagekey", required=False, help="Optional. Use this Azure Blob Storage account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--tenantid", required=False, help="Optional. Use this to define the Azure directory where to authenticate)")
parser.add_argument("--searchservice", help="Name of the Azure Cognitive Search service where content should be indexed (must exist already)")
parser.add_argument("--index", help="Name of the Azure Cognitive Search index where content should be indexed (will be created if it doesn't exist)")
parser.add_argument("--searchkey", required=False, help="Optional. Use this Azure Cognitive Search account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--remove", action="store_true", help="Remove references to this document from blob storage and the search index")
parser.add_argument("--removeall", action="store_true", help="Remove all blobs from blob storage and documents from the search index")
+ parser.add_argument("--localpdfparser", action="store_true", help="Use PyPdf local PDF parser (</s>
===========changed ref 3===========
# module: scripts.prepdocs
# offset: 1
<s> index")
+ parser.add_argument("--localpdfparser", action="store_true", help="Use PyPdf local PDF parser (supports only digital PDFs) instead of Azure Form Recognizer service to extract text, tables and layout from the documents")
+ parser.add_argument("--formrecognizerservice", required=False, help="Optional. Name of the Azure Form Recognizer service which will be used to extract text, tables and layout from the documents (must exist already)")
+ parser.add_argument("--formrecognizerkey", required=False, help="Optional. Use this Azure Form Recognizer account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--verbose", "-v", action="store_true", help="Verbose output")
args = parser.parse_args()
# Use the current user identity to connect to Azure services unless a key is explicitly set for any of them
azd_credential = AzureDeveloperCliCredential() if args.tenantid == None else AzureDeveloperCliCredential(tenant_id=args.tenantid)
default_creds = azd_credential if args.searchkey == None or args.storagekey == None else None
search_creds = default_creds if args.searchkey == None else AzureKeyCredential(args.searchkey)
if not args.skipblobs:
storage_creds = default_creds if args.storagekey == None else args.storagekey
+ if not args.localpdfparser:
+ # check if Azure Form Recognizer credentials are provided
+ if args.formrecognizerservice == None:
+ print("Error: Azure Form Recognizer service is not provided. Please provide formrecognizerservice or use --localpdfparser for local pypdf parser.")
+ exit(1)
+ formrecognizer_creds = default_creds if args.formrecognizerkey == None else AzureKeyCredential(args.formrecognizerkey)
|
scripts.prepdocs/create_sections
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6ac7c909c02d760bafd5e5e838fa8c2a46dd4aaf
|
Use Azure Form Recognizer as document preprocessing to extract text, tables, and document layout (#37)
|
<0>:<add> for i, (section, pagenum) in enumerate(split_text(page_map)):
<del> for i, (section, pagenum) in enumerate(split_text(pages)):
|
# module: scripts.prepdocs
+ def create_sections(filename, page_map):
- def create_sections(filename, pages):
<0> for i, (section, pagenum) in enumerate(split_text(pages)):
<1> yield {
<2> "id": re.sub("[^0-9a-zA-Z_-]","_",f"{filename}-{i}"),
<3> "content": section,
<4> "category": args.category,
<5> "sourcepage": blob_name_from_file_page(filename, pagenum),
<6> "sourcefile": filename
<7> }
<8>
|
===========unchanged ref 0===========
at: scripts.prepdocs.get_document_text
tables_on_page = [table for table in form_recognizer_results.tables if table.bounding_regions[0].page_number == page_num + 1]
===========changed ref 0===========
# module: scripts.prepdocs
+ def blob_name_from_file_page(filename, page = 0):
- def blob_name_from_file_page(filename, page):
+ if os.path.splitext(filename)[1].lower() == ".pdf":
+ return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
- return os.path.splitext(os.path.basename(filename))[0] + f"-{page}" + ".pdf"
+ else:
+ return os.path.basename(filename)
===========changed ref 1===========
# module: scripts.prepdocs
+ def table_to_html(table):
+ table_html = "<table>"
+ rows = [sorted([cell for cell in table.cells if cell.row_index == i], key=lambda cell: cell.column_index) for i in range(table.row_count)]
+ for row_cells in rows:
+ table_html += "<tr>"
+ for cell in row_cells:
+ tag = "th" if (cell.kind == "columnHeader" or cell.kind == "rowHeader") else "td"
+ cell_spans = ""
+ if cell.column_span > 1: cell_spans += f" colSpan={cell.column_span}"
+ if cell.row_span > 1: cell_spans += f" rowSpan={cell.row_span}"
+ table_html += f"<{tag}{cell_spans}>{html.escape(cell.content)}</{tag}>"
+ table_html +="</tr>"
+ table_html += "</table>"
+ return table_html
+
===========changed ref 2===========
# module: scripts.prepdocs
+ def upload_blobs(filename):
- def upload_blobs(pages):
blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
blob_container = blob_service.get_container_client(args.container)
if not blob_container.exists():
blob_container.create_container()
- for i in range(len(pages)):
- blob_name = blob_name_from_file_page(filename, i)
- if args.verbose: print(f"\tUploading blob for page {i} -> {blob_name}")
- f = io.BytesIO()
- writer = PdfWriter()
- writer.add_page(pages[i])
- writer.write(f)
- f.seek(0)
- blob_container.upload_blob(blob_name, f, overwrite=True)
===========changed ref 3===========
# module: scripts.prepdocs
+ def get_document_text(filename):
+ offset = 0
+ page_map = []
+ if args.localpdfparser:
+ reader = PdfReader(filename)
+ pages = reader.pages
+ for page_num, p in enumerate(pages):
+ page_text = p.extract_text()
+ page_map.append((page_num, offset, page_text))
+ offset += len(page_text)
+ else:
+ if args.verbose: print(f"Extracting text from '{filename}' using Azure Form Recognizer")
+ form_recognizer_client = DocumentAnalysisClient(endpoint=f"https://{args.formrecognizerservice}.cognitiveservices.azure.com/", credential=formrecognizer_creds, headers={"x-ms-useragent": "azure-search-chat-demo/1.0.0"})
+ with open(filename, "rb") as f:
+ poller = form_recognizer_client.begin_analyze_document("prebuilt-layout", document = f)
+ form_recognizer_results = poller.result()
+
+ for page_num, page in enumerate(form_recognizer_results.pages):
+ tables_on_page = [table for table in form_recognizer_results.tables if table.bounding_regions[0].page_number == page_num + 1]
+
+ # mark all positions of the table spans in the page
+ page_offset = page.spans[0].offset
+ page_length = page.spans[0].length
+ table_chars = [-1]*page_length
+ for table_id, table in enumerate(tables_on_page):
+ for span in table.spans:
+ # replace all table spans with "table_id" in table_chars array
+ for i in range(span.length):
+ idx = span.offset - page_offset + i
+ if idx >=0 and idx < page_length:
+ table_chars[idx] = table_id
+
+ </s>
===========changed ref 4===========
# module: scripts.prepdocs
+ def get_document_text(filename):
# offset: 1
<s> <add> if idx >=0 and idx < page_length:
+ table_chars[idx] = table_id
+
+ # build page text by replacing charcters in table spans with table html
+ page_text = ""
+ added_tables = set()
+ for idx, table_id in enumerate(table_chars):
+ if table_id == -1:
+ page_text += form_recognizer_results.content[page_offset + idx]
+ elif not table_id in added_tables:
+ page_text += table_to_html(tables_on_page[table_id])
+ added_tables.add(table_id)
+
+ page_text += " "
+ page_map.append((page_num, offset, page_text))
+ offset += len(page_text)
+
+ return page_map
+
===========changed ref 5===========
# module: scripts.prepdocs
+ def split_text(page_map):
- def split_text(pages):
SENTENCE_ENDINGS = [".", "!", "?"]
WORDS_BREAKS = [",", ";", ":", " ", "(", ")", "[", "]", "{", "}", "\t", "\n"]
if args.verbose: print(f"Splitting '{filename}' into sections")
-
- page_map = []
- offset = 0
- for i, p in enumerate(pages):
- text = p.extract_text()
- page_map.append((i, offset, text))
- offset += len(text)
def find_page(offset):
l = len(page_map)
for i in range(l - 1):
if offset >= page_map[i][1] and offset < page_map[i + 1][1]:
return i
return l - 1
all_text = "".join(p[2] for p in page_map)
length = len(all_text)
start = 0
end = length
while start + SECTION_OVERLAP < length:
last_word = -1
end = start + MAX_SECTION_LENGTH
if end > length:
end = length
else:
# Try to find the end of the sentence
while end < length and (end - start - MAX_SECTION_LENGTH) < SENTENCE_SEARCH_LIMIT and all_text[end] not in SENTENCE_ENDINGS:
if all_text[end] in WORDS_BREAKS:
last_word = end
end += 1
if end < length and all_text[end] not in SENTENCE_ENDINGS and last_word > 0:
end = last_word # Fall back to at least keeping a whole word
if end < length:
end += 1
# Try to find the start of the sentence or at least a whole word boundary
last_word = -1
while start > 0 and start > end - MAX_SECTION_LENGTH - 2 * SENTENCE_SEARCH_LIMIT and all</s>
|
app.backend.langchainadapters/HtmlCallbackHandler.on_tool_start
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
74d0776b127e955a739b53a8e459c4bd4085a581
|
Merge pull request #176 from mattmsft/mattmsft-patch-1
|
<1>:<add> pass
<del> self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
|
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
def on_tool_start(
self,
serialized: Dict[str, Any],
+ input_str: str,
- action: AgentAction,
color: Optional[str] = None,
**kwargs: Any,
) -> None:
<0> """Print out the log in specified color."""
<1> self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
<2>
|
===========unchanged ref 0===========
at: app.backend.langchainadapters.HtmlCallbackHandler
html: str = ""
at: typing
Dict = _alias(dict, 2, inst=False, name='Dict')
|
app.backend.lookuptool/CsvLookupTool.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
74d0776b127e955a739b53a8e459c4bd4085a581
|
Merge pull request #176 from mattmsft/mattmsft-patch-1
|
<0>:<add> super().__init__(name, self.lookup, description, callbacks=callbacks)
<del> super().__init__(name, self.lookup, description)
<1>:<del> self.data = {}
|
<s>
class CsvLookupTool(Tool):
+ def __init__(self, filename: path, key_field: str, name: str = "lookup",
+ description: str = "useful to look up details given an input key as opposite to searching data with an unstructured question",
- def __init__(self, filename: path, key_field: str, name: str = "lookup", description: str = "useful to look up details given an input key as opposite to searching data with an unstructured question"):
+ callbacks: Callbacks = None):
<0> super().__init__(name, self.lookup, description)
<1> self.data = {}
<2> with open(filename, newline='') as csvfile:
<3> reader = csv.DictReader(csvfile)
<4> for row in reader:
<5> self.data[row[key_field]] = "\n".join([f"{i}:{row[i]}" for i in row])
<6>
|
===========unchanged ref 0===========
at: app.backend.lookuptool.CsvLookupTool
lookup(key: str) -> Optional[str]
===========changed ref 0===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
+ def on_agent_action(
+ self,
+ action: AgentAction,
+ color: Optional[str] = None,
+ **kwargs: Any) -> Any:
+ self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
+
===========changed ref 1===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
def on_tool_start(
self,
serialized: Dict[str, Any],
+ input_str: str,
- action: AgentAction,
color: Optional[str] = None,
**kwargs: Any,
) -> None:
"""Print out the log in specified color."""
+ pass
- self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
|
app.backend.approaches.readretrieveread/ReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
74d0776b127e955a739b53a8e459c4bd4085a581
|
Merge pull request #176 from mattmsft/mattmsft-patch-1
|
<7>:<add> acs_tool = Tool(name="CognitiveSearch",
<add> func=lambda q: self.retrieve(q, overrides),
<add> description=self.CognitiveSearchToolDescription,
<add> callbacks=cb_manager)
<del> acs_tool = Tool(name = "CognitiveSearch", func = lambda q: self.retrieve(q, overrides), description = self.CognitiveSearchToolDescription)
<8>:<add> employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
<del> employee_tool = EmployeeInfoTool("Employee1")
|
<s> parts: first use GPT to see if we need more information,
# second if more data is needed use the requested "tool" to retrieve it. The last call to GPT answers the actual question.
# This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
# [1] E. Karpas, et al. arXiv:2205.00445
class ReadRetrieveReadApproach(Approach):
def run(self, q: str, overrides: dict) -> any:
<0> # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
<1> self.results = None
<2>
<3> # Use to capture thought process during iterations
<4> cb_handler = HtmlCallbackHandler()
<5> cb_manager = CallbackManager(handlers=[cb_handler])
<6>
<7> acs_tool = Tool(name = "CognitiveSearch", func = lambda q: self.retrieve(q, overrides), description = self.CognitiveSearchToolDescription)
<8> employee_tool = EmployeeInfoTool("Employee1")
<9> tools = [acs_tool, employee_tool]
<10>
<11> prompt = ZeroShotAgent.create_prompt(
<12> tools=tools,
<13> prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
<14> suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
<15> input_variables = ["input", "agent_scratchpad"])
<16> llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
<17> chain = LLMChain(llm = llm, prompt = prompt)
<18> agent_exec = AgentExecutor.from_agent_and_tools(
<19> agent = ZeroShotAgent(llm_chain = chain, tools = tools),
<20> tools = tools,
<21> verbose = True,
<22> callback_manager = cb_manager)
<23> result =</s>
|
===========below chunk 0===========
<s>PT to see if we need more information,
# second if more data is needed use the requested "tool" to retrieve it. The last call to GPT answers the actual question.
# This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
# [1] E. Karpas, et al. arXiv:2205.00445
class ReadRetrieveReadApproach(Approach):
def run(self, q: str, overrides: dict) -> any:
# offset: 1
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
===========unchanged ref 0===========
at: app.backend.approaches.readretrieveread
EmployeeInfoTool(employee_name: str, callbacks: Callbacks=None)
at: app.backend.approaches.readretrieveread.ReadRetrieveReadApproach
template_prefix = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " \
"Answer the question using only the data provided in the information sources below. " \
"For tabular information return it as an html table. Do not return markdown format. " \
"Each source has a name followed by colon and the actual data, quote the source name for each piece of data you use in the response. " \
"For example, if the question is \"What color is the sky?\" and one of the information sources says \"info123: the sky is blue whenever it's not cloudy\", then answer with \"The sky is blue [info123]\" " \
"It's important to strictly follow the format where the name of the source is in square brackets at the end of the sentence, and only up to the prefix before the colon (\":\"). " \
"If there are multiple sources, cite each one in their own square brackets. For example, use \"[info343][ref-76]\" and not \"[info343,ref-76]\". " \
"Never quote tool names as sources." \
"If you cannot answer using the sources below, say that you don't know. " \
"\n\nYou can access to the following tools:"
template_suffix = """
Begin!
Question: {input}
Thought: {agent_scratchpad}"""
CognitiveSearchToolDescription = "useful for searching the Microsoft employee benefits information such as healthcare plans, retirement plans, etc."
retrieve(q: str, overrides: dict) -> any
at: app.backend.approaches.readretrieveread.ReadRetrieveReadApproach.__init__
self.openai_deployment = openai_deployment
===========unchanged ref 1===========
at: app.backend.approaches.readretrieveread.ReadRetrieveReadApproach.retrieve
self.results = [doc[self.sourcepage_field] + ":" + nonewlines(" -.- ".join([c.text for c in doc['@search.captions']])) for doc in r]
self.results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:250]) for doc in r]
at: approaches.approach.Approach
run(self, q: str, use_summaries: bool) -> any
at: openai
api_key = os.environ.get("OPENAI_API_KEY")
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
+ def on_agent_action(
+ self,
+ action: AgentAction,
+ color: Optional[str] = None,
+ **kwargs: Any) -> Any:
+ self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
+
===========changed ref 1===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
def on_tool_start(
self,
serialized: Dict[str, Any],
+ input_str: str,
- action: AgentAction,
color: Optional[str] = None,
**kwargs: Any,
) -> None:
"""Print out the log in specified color."""
+ pass
- self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
===========changed ref 2===========
<s>
class CsvLookupTool(Tool):
+ def __init__(self, filename: path, key_field: str, name: str = "lookup",
+ description: str = "useful to look up details given an input key as opposite to searching data with an unstructured question",
- def __init__(self, filename: path, key_field: str, name: str = "lookup", description: str = "useful to look up details given an input key as opposite to searching data with an unstructured question"):
+ callbacks: Callbacks = None):
+ super().__init__(name, self.lookup, description, callbacks=callbacks)
- super().__init__(name, self.lookup, description)
- self.data = {}
with open(filename, newline='') as csvfile:
reader = csv.DictReader(csvfile)
for row in reader:
self.data[row[key_field]] = "\n".join([f"{i}:{row[i]}" for i in row])
|
app.backend.approaches.readretrieveread/EmployeeInfoTool.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
74d0776b127e955a739b53a8e459c4bd4085a581
|
Merge pull request #176 from mattmsft/mattmsft-patch-1
|
<0>:<add> super().__init__(filename="data/employeeinfo.csv",
<add> key_field="name",
<add> name="Employee",
<add> description="useful for answering questions about the employee, their benefits and other personal information",
<add> callbacks=callbacks)
<del> super().__init__(filename = "data/employeeinfo.csv", key_field = "name", name = "Employee", description = "useful for answering questions about the employee, their benefits and other personal information")
|
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
+ def __init__(self, employee_name: str, callbacks: Callbacks = None):
- def __init__(self, employee_name: str):
<0> super().__init__(filename = "data/employeeinfo.csv", key_field = "name", name = "Employee", description = "useful for answering questions about the employee, their benefits and other personal information")
<1> self.func = self.employee_info
<2> self.employee_name = employee_name
<3>
|
===========unchanged ref 0===========
at: lookuptool
CsvLookupTool(filename: path, key_field: str, name: str="lookup", description: str="useful to look up details given an input key as opposite to searching data with an unstructured question", callbacks: Callbacks=None)
at: lookuptool.CsvLookupTool
data: dict[str, str] = {}
__init__(self, filename: path, key_field: str, name: str="lookup", description: str="useful to look up details given an input key as opposite to searching data with an unstructured question", callbacks: Callbacks=None)
===========changed ref 0===========
<s> parts: first use GPT to see if we need more information,
# second if more data is needed use the requested "tool" to retrieve it. The last call to GPT answers the actual question.
# This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
# [1] E. Karpas, et al. arXiv:2205.00445
class ReadRetrieveReadApproach(Approach):
def run(self, q: str, overrides: dict) -> any:
# Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
self.results = None
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
+ acs_tool = Tool(name="CognitiveSearch",
+ func=lambda q: self.retrieve(q, overrides),
+ description=self.CognitiveSearchToolDescription,
+ callbacks=cb_manager)
- acs_tool = Tool(name = "CognitiveSearch", func = lambda q: self.retrieve(q, overrides), description = self.CognitiveSearchToolDescription)
+ employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
- employee_tool = EmployeeInfoTool("Employee1")
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec =</s>
===========changed ref 1===========
<s>PT to see if we need more information,
# second if more data is needed use the requested "tool" to retrieve it. The last call to GPT answers the actual question.
# This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
# [1] E. Karpas, et al. arXiv:2205.00445
class ReadRetrieveReadApproach(Approach):
def run(self, q: str, overrides: dict) -> any:
# offset: 1
<s>ai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
agent = ZeroShotAgent(llm_chain = chain, tools = tools),
tools = tools,
verbose = True,
callback_manager = cb_manager)
result = agent_exec.run(q)
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
===========changed ref 2===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
+ def on_agent_action(
+ self,
+ action: AgentAction,
+ color: Optional[str] = None,
+ **kwargs: Any) -> Any:
+ self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
+
===========changed ref 3===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
def on_tool_start(
self,
serialized: Dict[str, Any],
+ input_str: str,
- action: AgentAction,
color: Optional[str] = None,
**kwargs: Any,
) -> None:
"""Print out the log in specified color."""
+ pass
- self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
===========changed ref 4===========
<s>
class CsvLookupTool(Tool):
+ def __init__(self, filename: path, key_field: str, name: str = "lookup",
+ description: str = "useful to look up details given an input key as opposite to searching data with an unstructured question",
- def __init__(self, filename: path, key_field: str, name: str = "lookup", description: str = "useful to look up details given an input key as opposite to searching data with an unstructured question"):
+ callbacks: Callbacks = None):
+ super().__init__(name, self.lookup, description, callbacks=callbacks)
- super().__init__(name, self.lookup, description)
- self.data = {}
with open(filename, newline='') as csvfile:
reader = csv.DictReader(csvfile)
for row in reader:
self.data[row[key_field]] = "\n".join([f"{i}:{row[i]}" for i in row])
|
app.backend.approaches.readdecomposeask/ReadDecomposeAsk.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
74d0776b127e955a739b53a8e459c4bd4085a581
|
Merge pull request #176 from mattmsft/mattmsft-patch-1
|
<9>:<add> Tool(name="Search", func=lambda q: self.search(q, overrides), description="useful for when you need to ask with search", callbacks=cb_manager),
<add> Tool(name="Lookup", func=self.lookup, description="useful for when you need to ask with lookup", callbacks=cb_manager)
<del> Tool(name="Search", func=lambda q: self.search(q, overrides)),
<10>:<del> Tool(name="Lookup", func=self.lookup)
<23>:<add> # Replace substrings of the form <file.ext> with [file.ext] so that the frontend can render them as links, match them with a regex to avoid
<add> # generalizing too much and disrupt HTML snippets if present
<add> result = re.sub(r"<([a-zA-Z0-9_ \-\.]+)>", r"[\1]", result)
<del> # Fix up references to they look like what the frontend expects ([] instead of ()), need a better citation format since parentheses are so common
<24>:<del> result = result.replace("(", "[").replace(")", "]")
|
# module: app.backend.approaches.readdecomposeask
class ReadDecomposeAsk(Approach):
def run(self, q: str, overrides: dict) -> any:
<0> # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
<1> self.results = None
<2>
<3> # Use to capture thought process during iterations
<4> cb_handler = HtmlCallbackHandler()
<5> cb_manager = CallbackManager(handlers=[cb_handler])
<6>
<7> llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
<8> tools = [
<9> Tool(name="Search", func=lambda q: self.search(q, overrides)),
<10> Tool(name="Lookup", func=self.lookup)
<11> ]
<12>
<13> # Like results above, not great to keep this as a global, will interfere with interleaving
<14> global prompt
<15> prompt_prefix = overrides.get("prompt_template")
<16> prompt = PromptTemplate.from_examples(
<17> EXAMPLES, SUFFIX, ["input", "agent_scratchpad"], prompt_prefix + "\n\n" + PREFIX if prompt_prefix else PREFIX)
<18>
<19> agent = ReAct.from_llm_and_tools(llm, tools)
<20> chain = AgentExecutor.from_agent_and_tools(agent, tools, verbose=True, callback_manager=cb_manager)
<21> result = chain.run(q)
<22>
<23> # Fix up references to they look like what the frontend expects ([] instead of ()), need a better citation format since parentheses are so common
<24> result = result.replace("(", "[").replace(")", "]")
<25>
<26> return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
<27>
|
===========unchanged ref 0===========
at: app.backend.approaches.readdecomposeask
EXAMPLES = [
"""Question: What is the elevation range for the area that the eastern sector of the
Colorado orogeny extends into?
Thought: I need to search Colorado orogeny, find the area that the eastern sector
of the Colorado orogeny extends into, then find the elevation range of the
area.
Action: Search[Colorado orogeny]
Observation: <info1.pdf> The Colorado orogeny was an episode of mountain building (an orogeny) in
Colorado and surrounding areas.
Thought: It does not mention the eastern sector. So I need to look up eastern
sector.
Action: Lookup[eastern sector]
Observation: <info2.txt> (Result 1 / 1) The eastern sector extends into the High Plains and is called
the Central Plains orogeny.
Thought: The eastern sector of Colorado orogeny extends into the High Plains. So I
need to search High Plains and find its elevation range.
Action: Search[High Plains]
Observation: <some_file.pdf> High Plains refers to one of two distinct land regions
Thought: I need to instead search High Plains (United States).
Action: Search[High Plains (United States)]
Observation: <filea.pdf> The High Plains are a subregion of the Great Plains. <another-ref.docx> From east to west, the
High Plains rise in elevation from around 1,800 to 7,000 ft (550 to 2,130
m).
Thought: High Plains rise in elevation from around 1,800 to 7,000 ft, so the answer
is 1,800 to 7,000 ft.
Action: Finish[1,800 to 7,000 ft <filea.pdf>]""",
"""Question: Musician and satirist Allie Goertz wrote a song about the "The Simpsons"
character Milhouse, who</s>
===========unchanged ref 1===========
SUFFIX = """\nQuestion: {input}
{agent_scratchpad}"""
PREFIX = "Answer questions as shown in the following examples, by splitting the question into individual search or lookup actions to find facts until you can answer the question. " \
"Observations are prefixed by their source name in angled brackets, source names MUST be included with the actions in the answers." \
"All questions must be answered from the results from search or look up actions, only facts resulting from those can be used in an answer. "
at: app.backend.approaches.readdecomposeask.ReadDecomposeAsk
search(q: str, overrides: dict) -> str
lookup(q: str) -> str
at: app.backend.approaches.readdecomposeask.ReadDecomposeAsk.__init__
self.openai_deployment = openai_deployment
at: app.backend.approaches.readdecomposeask.ReadDecomposeAsk.search
self.results = [doc[self.sourcepage_field] + ":" + nonewlines(" . ".join([c.text for c in doc['@search.captions'] ])) for doc in r]
self.results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:500]) for doc in r]
at: approaches.approach.Approach
run(self, q: str, use_summaries: bool) -> any
at: openai
api_key = os.environ.get("OPENAI_API_KEY")
===========unchanged ref 2===========
at: re
sub(pattern: AnyStr, repl: AnyStr, string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: Pattern[AnyStr], repl: Callable[[Match[AnyStr]], AnyStr], string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: AnyStr, repl: Callable[[Match[AnyStr]], AnyStr], string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
sub(pattern: Pattern[AnyStr], repl: AnyStr, string: AnyStr, count: int=..., flags: _FlagsType=...) -> AnyStr
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
+ def on_agent_action(
+ self,
+ action: AgentAction,
+ color: Optional[str] = None,
+ **kwargs: Any) -> Any:
+ self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
+
===========changed ref 1===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
def on_tool_start(
self,
serialized: Dict[str, Any],
+ input_str: str,
- action: AgentAction,
color: Optional[str] = None,
**kwargs: Any,
) -> None:
"""Print out the log in specified color."""
+ pass
- self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
===========changed ref 2===========
<s>
class CsvLookupTool(Tool):
+ def __init__(self, filename: path, key_field: str, name: str = "lookup",
+ description: str = "useful to look up details given an input key as opposite to searching data with an unstructured question",
- def __init__(self, filename: path, key_field: str, name: str = "lookup", description: str = "useful to look up details given an input key as opposite to searching data with an unstructured question"):
+ callbacks: Callbacks = None):
+ super().__init__(name, self.lookup, description, callbacks=callbacks)
- super().__init__(name, self.lookup, description)
- self.data = {}
with open(filename, newline='') as csvfile:
reader = csv.DictReader(csvfile)
for row in reader:
self.data[row[key_field]] = "\n".join([f"{i}:{row[i]}" for i in row])
===========changed ref 3===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
+ def __init__(self, employee_name: str, callbacks: Callbacks = None):
- def __init__(self, employee_name: str):
+ super().__init__(filename="data/employeeinfo.csv",
+ key_field="name",
+ name="Employee",
+ description="useful for answering questions about the employee, their benefits and other personal information",
+ callbacks=callbacks)
- super().__init__(filename = "data/employeeinfo.csv", key_field = "name", name = "Employee", description = "useful for answering questions about the employee, their benefits and other personal information")
self.func = self.employee_info
self.employee_name = employee_name
|
app.backend.approaches.readretrieveread/EmployeeInfoTool.employee_info
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
46e75e16d88a3f106f3e32bbb75d65617d0b83e3
|
Merge pull request #345 from pamelafox/other-way
|
<0>:<add> return self.lookup(name)
<del> return self.lookup(self.employee_name)
|
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
+ def employee_info(self, name: str) -> str:
- def employee_info(self, unused: str) -> str:
<0> return self.lookup(self.employee_name)
<1>
|
===========changed ref 0===========
<s> information
- # is present then formulate an answer. Each iteration consists of two parts: first use GPT to see if we need more information,
- # second if more data is needed use the requested "tool" to retrieve it. The last call to GPT answers the actual question.
- # This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
- # [1] E. Karpas, et al. arXiv:2205.00445
class ReadRetrieveReadApproach(Approach):
+ """
+ Attempt to answer questions by iteratively evaluating the question to see what information is missing, and once all information
+ is present then formulate an answer. Each iteration consists of two parts:
+ 1. use GPT to see if we need more information
+ 2. if more data is needed, use the requested "tool" to retrieve it.
+ The last call to GPT answers the actual question.
+ This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
+
+ [1] E. Karpas, et al. arXiv:2205.00445
+ """
+
template_prefix = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " \
"Answer the question using only the data provided in the information sources below. " \
"For tabular information return it as an html table. Do not return markdown format. " \
"Each source has a name followed by colon and the actual data, quote the source name for each piece of data you use in the response. " \
"For example, if the question is \"What color is the sky?\" and one of the information sources says \"info123: the sky is blue whenever it's not cloudy\", then answer with \"The sky is blue [info123]\" " \
"It's important to strictly follow the format where the name of the source is in square brackets at the end of the sentence, and only up to the prefix before the colon (\":\"). " \
"If there are multiple sources, cite each one in their own square brackets. For example</s>
===========changed ref 1===========
<s> present then formulate an answer. Each iteration consists of two parts: first use GPT to see if we need more information,
- # second if more data is needed use the requested "tool" to retrieve it. The last call to GPT answers the actual question.
- # This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
- # [1] E. Karpas, et al. arXiv:2205.00445
class ReadRetrieveReadApproach(Approach):
# offset: 1
<s> prefix before the colon (\":\"). " \
"If there are multiple sources, cite each one in their own square brackets. For example, use \"[info343][ref-76]\" and not \"[info343,ref-76]\". " \
"Never quote tool names as sources." \
"If you cannot answer using the sources below, say that you don't know. " \
"\n\nYou can access to the following tools:"
template_suffix = """
Begin!
Question: {input}
Thought: {agent_scratchpad}"""
CognitiveSearchToolDescription = "useful for searching the Microsoft employee benefits information such as healthcare plans, retirement plans, etc."
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_chat_history_as_text
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
46e75e16d88a3f106f3e32bbb75d65617d0b83e3
|
Merge pull request #345 from pamelafox/other-way
|
<2>:<add> history_text = """<|im_start|>user""" + "\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot", "") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
<del> history_text = """<|im_start|>user""" +"\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
|
<s> a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class ChatReadRetrieveReadApproach(Approach):
+ def get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str:
- def get_chat_history_as_text(self, history, include_last_turn=True, approx_max_tokens=1000) -> str:
<0> history_text = ""
<1> for h in reversed(history if include_last_turn else history[:-1]):
<2> history_text = """<|im_start|>user""" +"\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
<3> if len(history_text) > approx_max_tokens*4:
<4> break
<5> return history_text
<6>
|
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
prompt_prefix = """<|im_start|>system
Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
Sources:
{sources}
<|im_end|>
{chat_history}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.run
prompt = self.prompt_prefix.format(injected_prompt=prompt_override[3:] + "\n", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
prompt = prompt_override.format(sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
q = completion.choices[0].text
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
===========unchanged ref 2===========
completion = openai.Completion.create(
engine=self.chatgpt_deployment,
prompt=prompt,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1,
stop=["<|im_end|>", "<|im_start|>"])
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
- # Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
- # top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class ChatReadRetrieveReadApproach(Approach):
+ """
+ Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
+ top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
+ (answer) with that prompt.
+ """
+
prompt_prefix = """<|im_start|>system
Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
+ Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
- Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brakets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
Sources:
{sources}
<|im_end|>
{chat_history}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
</s>
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
- # Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
- # top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class ChatReadRetrieveReadApproach(Approach):
# offset: 1
<s>Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
===========changed ref 2===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
+ def employee_info(self, name: str) -> str:
- def employee_info(self, unused: str) -> str:
+ return self.lookup(name)
- return self.lookup(self.employee_name)
|
app.backend.app/content_file
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
46e75e16d88a3f106f3e32bbb75d65617d0b83e3
|
Merge pull request #345 from pamelafox/other-way
|
<1>:<add> if not blob.properties or not blob.properties.has_key("content_settings"):
<add> abort(404)
<4>:<add> blob_file = io.BytesIO()
<add> blob.readinto(blob_file)
<add> blob_file.seek(0)
<add> return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
<del> return blob.readall(), 200, {"Content-Type": mime_type, "Content-Disposition": f"inline; filename={path}"}
|
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
@app.route("/content/<path>")
def content_file(path):
<0> blob = blob_container.get_blob_client(path).download_blob()
<1> mime_type = blob.properties["content_settings"]["content_type"]
<2> if mime_type == "application/octet-stream":
<3> mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
<4> return blob.readall(), 200, {"Content-Type": mime_type, "Content-Disposition": f"inline; filename={path}"}
<5>
|
===========unchanged ref 0===========
at: app.backend.app
blob_container = blob_client.get_container_client(AZURE_STORAGE_CONTAINER)
app = Flask(__name__)
===========changed ref 0===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
+ def employee_info(self, name: str) -> str:
- def employee_info(self, unused: str) -> str:
+ return self.lookup(name)
- return self.lookup(self.employee_name)
===========changed ref 1===========
<s> a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class ChatReadRetrieveReadApproach(Approach):
+ def get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str:
- def get_chat_history_as_text(self, history, include_last_turn=True, approx_max_tokens=1000) -> str:
history_text = ""
for h in reversed(history if include_last_turn else history[:-1]):
+ history_text = """<|im_start|>user""" + "\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot", "") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
- history_text = """<|im_start|>user""" +"\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
if len(history_text) > approx_max_tokens*4:
break
return history_text
===========changed ref 2===========
# module: app.backend.approaches.retrievethenread
- # Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
- # top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class RetrieveThenReadApproach(Approach):
+ """
+ Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
+ top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
+ (answer) with that prompt.
+ """
+
template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
"If you cannot answer using the sources below, say you don't know. " + \
"""
###
Question: 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
Answer:
In-network deductibles are $500 for employee and $</s>
===========changed ref 3===========
# module: app.backend.approaches.retrievethenread
- # Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
- # top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class RetrieveThenReadApproach(Approach):
# offset: 1
<s>ake, Swedish and others in the region
Answer:
In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf].
###
Question: '{q}'?
Sources:
{retrieved}
Answer:
"""
===========changed ref 4===========
<s> information
- # is present then formulate an answer. Each iteration consists of two parts: first use GPT to see if we need more information,
- # second if more data is needed use the requested "tool" to retrieve it. The last call to GPT answers the actual question.
- # This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
- # [1] E. Karpas, et al. arXiv:2205.00445
class ReadRetrieveReadApproach(Approach):
+ """
+ Attempt to answer questions by iteratively evaluating the question to see what information is missing, and once all information
+ is present then formulate an answer. Each iteration consists of two parts:
+ 1. use GPT to see if we need more information
+ 2. if more data is needed, use the requested "tool" to retrieve it.
+ The last call to GPT answers the actual question.
+ This is inspired by the MKRL paper[1] and applied here using the implementation in Langchain.
+
+ [1] E. Karpas, et al. arXiv:2205.00445
+ """
+
template_prefix = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " \
"Answer the question using only the data provided in the information sources below. " \
"For tabular information return it as an html table. Do not return markdown format. " \
"Each source has a name followed by colon and the actual data, quote the source name for each piece of data you use in the response. " \
"For example, if the question is \"What color is the sky?\" and one of the information sources says \"info123: the sky is blue whenever it's not cloudy\", then answer with \"The sky is blue [info123]\" " \
"It's important to strictly follow the format where the name of the source is in square brackets at the end of the sentence, and only up to the prefix before the colon (\":\"). " \
"If there are multiple sources, cite each one in their own square brackets. For example</s>
|
app.backend.app/ask
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
46e75e16d88a3f106f3e32bbb75d65617d0b83e3
|
Merge pull request #345 from pamelafox/other-way
|
<1>:<add> if not request.json:
<add> return jsonify({"error": "request must be json"}), 400
|
# module: app.backend.app
@app.route("/ask", methods=["POST"])
def ask():
<0> ensure_openai_token()
<1> approach = request.json["approach"]
<2> try:
<3> impl = ask_approaches.get(approach)
<4> if not impl:
<5> return jsonify({"error": "unknown approach"}), 400
<6> r = impl.run(request.json["question"], request.json.get("overrides") or {})
<7> return jsonify(r)
<8> except Exception as e:
<9> logging.exception("Exception in /ask")
<10> return jsonify({"error": str(e)}), 500
<11>
|
===========unchanged ref 0===========
at: app.backend.app
app = Flask(__name__)
ensure_openai_token()
at: app.backend.app.content_file
blob = blob_container.get_blob_client(path).download_blob()
at: io
BytesIO(initial_bytes: bytes=...)
at: io.BytesIO
seek(self, offset: int, whence: int=..., /) -> int
at: mimetypes
guess_type(url: Union[Text, PathLike[str]], strict: bool=...) -> Tuple[Optional[str], Optional[str]]
===========changed ref 0===========
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
@app.route("/content/<path>")
def content_file(path):
blob = blob_container.get_blob_client(path).download_blob()
+ if not blob.properties or not blob.properties.has_key("content_settings"):
+ abort(404)
mime_type = blob.properties["content_settings"]["content_type"]
if mime_type == "application/octet-stream":
mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
+ blob_file = io.BytesIO()
+ blob.readinto(blob_file)
+ blob_file.seek(0)
+ return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
- return blob.readall(), 200, {"Content-Type": mime_type, "Content-Disposition": f"inline; filename={path}"}
===========changed ref 1===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
+ def employee_info(self, name: str) -> str:
- def employee_info(self, unused: str) -> str:
+ return self.lookup(name)
- return self.lookup(self.employee_name)
===========changed ref 2===========
<s> a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class ChatReadRetrieveReadApproach(Approach):
+ def get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str:
- def get_chat_history_as_text(self, history, include_last_turn=True, approx_max_tokens=1000) -> str:
history_text = ""
for h in reversed(history if include_last_turn else history[:-1]):
+ history_text = """<|im_start|>user""" + "\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot", "") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
- history_text = """<|im_start|>user""" +"\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
if len(history_text) > approx_max_tokens*4:
break
return history_text
===========changed ref 3===========
# module: app.backend.approaches.retrievethenread
- # Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
- # top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class RetrieveThenReadApproach(Approach):
+ """
+ Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
+ top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
+ (answer) with that prompt.
+ """
+
template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
"If you cannot answer using the sources below, say you don't know. " + \
"""
###
Question: 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
Answer:
In-network deductibles are $500 for employee and $</s>
===========changed ref 4===========
# module: app.backend.approaches.retrievethenread
- # Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
- # top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
- # (answer) with that prompt.
class RetrieveThenReadApproach(Approach):
# offset: 1
<s>ake, Swedish and others in the region
Answer:
In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf].
###
Question: '{q}'?
Sources:
{retrieved}
Answer:
"""
|
app.backend.app/chat
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
46e75e16d88a3f106f3e32bbb75d65617d0b83e3
|
Merge pull request #345 from pamelafox/other-way
|
<1>:<add> if not request.json:
<add> return jsonify({"error": "request must be json"}), 400
|
# module: app.backend.app
@app.route("/chat", methods=["POST"])
def chat():
<0> ensure_openai_token()
<1> approach = request.json["approach"]
<2> try:
<3> impl = chat_approaches.get(approach)
<4> if not impl:
<5> return jsonify({"error": "unknown approach"}), 400
<6> r = impl.run(request.json["history"], request.json.get("overrides") or {})
<7> return jsonify(r)
<8> except Exception as e:
<9> logging.exception("Exception in /chat")
<10> return jsonify({"error": str(e)}), 500
<11>
|
===========unchanged ref 0===========
at: app.backend.app
ask_approaches = {
"rtr": RetrieveThenReadApproach(search_client, AZURE_OPENAI_GPT_DEPLOYMENT, KB_FIELDS_SOURCEPAGE, KB_FIELDS_CONTENT),
"rrr": ReadRetrieveReadApproach(search_client, AZURE_OPENAI_GPT_DEPLOYMENT, KB_FIELDS_SOURCEPAGE, KB_FIELDS_CONTENT),
"rda": ReadDecomposeAsk(search_client, AZURE_OPENAI_GPT_DEPLOYMENT, KB_FIELDS_SOURCEPAGE, KB_FIELDS_CONTENT)
}
app = Flask(__name__)
ensure_openai_token()
at: app.backend.app.ask
approach = request.json["approach"]
at: approaches.readdecomposeask.ReadDecomposeAsk
run(q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 1===========
at: approaches.readretrieveread.ReadRetrieveReadApproach
template_prefix = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " \
"Answer the question using only the data provided in the information sources below. " \
"For tabular information return it as an html table. Do not return markdown format. " \
"Each source has a name followed by colon and the actual data, quote the source name for each piece of data you use in the response. " \
"For example, if the question is \"What color is the sky?\" and one of the information sources says \"info123: the sky is blue whenever it's not cloudy\", then answer with \"The sky is blue [info123]\" " \
"It's important to strictly follow the format where the name of the source is in square brackets at the end of the sentence, and only up to the prefix before the colon (\":\"). " \
"If there are multiple sources, cite each one in their own square brackets. For example, use \"[info343][ref-76]\" and not \"[info343,ref-76]\". " \
"Never quote tool names as sources." \
"If you cannot answer using the sources below, say that you don't know. " \
"\n\nYou can access to the following tools:"
template_suffix = """
Begin!
Question: {input}
Thought: {agent_scratchpad}"""
CognitiveSearchToolDescription = "useful for searching the Microsoft employee benefits information such as healthcare plans, retirement plans, etc."
run(q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 2===========
at: approaches.retrievethenread.RetrieveThenReadApproach
template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
"If you cannot answer using the sources below, say you don't know. " + \
"""
###
Question: 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
Answer:
In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf].
###
Question: '{q}'?
Sources:
{retrieved}
Answer:
"""
run(q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 3===========
at: logging
exception(msg: Any, *args: Any, exc_info: _ExcInfoType=..., stack_info: bool=..., extra: Optional[Dict[str, Any]]=..., **kwargs: Any) -> None
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.app
@app.route("/ask", methods=["POST"])
def ask():
ensure_openai_token()
+ if not request.json:
+ return jsonify({"error": "request must be json"}), 400
approach = request.json["approach"]
try:
impl = ask_approaches.get(approach)
if not impl:
return jsonify({"error": "unknown approach"}), 400
r = impl.run(request.json["question"], request.json.get("overrides") or {})
return jsonify(r)
except Exception as e:
logging.exception("Exception in /ask")
return jsonify({"error": str(e)}), 500
===========changed ref 1===========
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
@app.route("/content/<path>")
def content_file(path):
blob = blob_container.get_blob_client(path).download_blob()
+ if not blob.properties or not blob.properties.has_key("content_settings"):
+ abort(404)
mime_type = blob.properties["content_settings"]["content_type"]
if mime_type == "application/octet-stream":
mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
+ blob_file = io.BytesIO()
+ blob.readinto(blob_file)
+ blob_file.seek(0)
+ return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
- return blob.readall(), 200, {"Content-Type": mime_type, "Content-Disposition": f"inline; filename={path}"}
===========changed ref 2===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
+ def employee_info(self, name: str) -> str:
- def employee_info(self, unused: str) -> str:
+ return self.lookup(name)
- return self.lookup(self.employee_name)
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
185b0c64d9ae748e241ca253e1d681284849227a
|
upgrade to chat completion api
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
if prompt_override is None:
prompt = self.prompt_prefix.format(injected_prompt="", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
prompt = self.prompt_prefix.format(injected_prompt=prompt_override[3:] + "\n", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
else:
prompt = prompt_override.format(sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
completion = openai.Completion.create(
engine=self.chatgpt_deployment,
prompt=prompt,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1,
stop=["<|im_end|>", "<|im_start|>"])
return {"data_points": results, "answer": completion.choices[0].text, "tho</s>
===========below chunk 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 2
<s>_start|>"])
return {"data_points": results, "answer": completion.choices[0].text, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
prompt_prefix = """<|im_start|>system
Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
Sources:
{sources}
<|im_end|>
{chat_history}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
===========unchanged ref 2===========
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
b21d06f347d144171f432a36f60cd4b9b4106054
|
update model version
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
if prompt_override is None:
prompt = self.prompt_prefix.format(injected_prompt="", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
prompt = self.prompt_prefix.format(injected_prompt=prompt_override[3:] + "\n", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
else:
prompt = prompt_override.format(sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
print("history: \n")
print(history)
print("prompt: \n")
print(prompt)
messages = self.get_messages_from_prompt(prompt)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model="gpt-3.5-turbo",
messages=messages,
temperature=overrides.get("temperature") or 0.7</s>
===========below chunk 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 2
<s>pt-3.5-turbo",
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1,
stop=["<|im_end|>", "<|im_start|>"])
chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chatContent, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
prompt_prefix = """<|im_start|>system
Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
Sources:
{sources}
<|im_end|>
{chat_history}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_messages_from_prompt(prompt: str) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
===========unchanged ref 2===========
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
695d6901b06911771cd798ef9017afa373c05ca6
|
add message construction and validate tokens
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
if prompt_override is None:
prompt = self.prompt_prefix.format(injected_prompt="", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
prompt = self.prompt_prefix.format(injected_prompt=prompt_override[3:] + "\n", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
else:
prompt = prompt_override.format(sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
messages = self.get_messages_from_prompt(prompt)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model="gpt-3.5-turbo",
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1,
stop=["<|im_end|>"</s>
===========below chunk 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 2
<s>
max_tokens=1024,
n=1,
stop=["<|im_end|>", "<|im_start|>"])
chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chatContent, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
prompt_prefix = """<|im_start|>system
Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
Sources:
{sources}
<|im_end|>
{chat_history}
"""
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of Sources:. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
===========unchanged ref 1===========
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
===========unchanged ref 2===========
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
e0db5fefd572d617b0ad7f7b746d25bdfa78bfa5
|
Address comments
|
<2>:<add> self.chatgpt_model = chatgpt_model
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
<0> self.search_client = search_client
<1> self.chatgpt_deployment = chatgpt_deployment
<2> self.gpt_deployment = gpt_deployment
<3> self.sourcepage_field = sourcepage_field
<4> self.content_field = content_field
<5>
|
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of Sources:. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.gpt_deployment = gpt_deployment
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.run
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
===========unchanged ref 2===========
prompt = self.prompt_prefix.format(injected_prompt="", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
prompt = self.prompt_prefix.format(injected_prompt=prompt_override[3:] + "\n", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
prompt = prompt_override.format(sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
completion = openai.Completion.create(
engine=self.gpt_deployment,
prompt=prompt,
temperature=0.0,
max_tokens=32,
n=1,
stop=["\n"])
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
#Chat roles
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
- prompt_prefix = """<|im_start|>system
- Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
- Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
- For tabular information return it as an html table. Do not return markdown format.
- Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
- {follow_up_questions_prompt}
- {injected_prompt}
- Sources:
- {sources}
- <|im_end|>
- {chat_history}
- """
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of Sources:. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use</s>
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
e0db5fefd572d617b0ad7f7b746d25bdfa78bfa5
|
Address comments
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
if prompt_override is None:
prompt = self.prompt_prefix.format(injected_prompt="", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
prompt = self.prompt_prefix.format(injected_prompt=prompt_override[3:] + "\n", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
else:
prompt = prompt_override.format(sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
print(messages)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model="gpt-3.5-turbo",
messages=messages,
temperature=overrides.get("</s>
===========below chunk 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 2
<s>,
model="gpt-3.5-turbo",
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chatContent, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
prompt_prefix = """<|im_start|>system
Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
Sources:
{sources}
<|im_end|>
{chat_history}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
===========unchanged ref 1===========
get_messages_from_history(prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
+ self.chatgpt_model = chatgpt_model
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
|
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_messages_from_history
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
e0db5fefd572d617b0ad7f7b746d25bdfa78bfa5
|
Address comments
|
<10>:<add> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
<del> token_count = token_count + self.num_tokens_from_messages(messages, "gpt-3.5-turbo")
<15>:<add> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
<del> token_count = token_count + self.num_tokens_from_messages(messages, "gpt-3.5-turbo")
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
<0> messages = []
<1> token_count = 0
<2> if prompt_override is None:
<3> system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
<4> elif prompt_override.startswith(">>>"):
<5> system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
<6> else:
<7> system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
<8>
<9> messages.append({"role":self.SYSTEM, "content": system_message})
<10> token_count = token_count + self.num_tokens_from_messages(messages, "gpt-3.5-turbo")
<11>
<12> #latest conversation
<13> userContent = history[-1]["user"] + "\nSources:" + sources
<14> messages.append({"role": self.USER, "content": userContent})
<15> token_count = token_count + self.num_tokens_from_messages(messages, "gpt-3.5-turbo")
<16>
<17> '''
<18> Enqueue in reverse order
<19> if limit exceeds truncate old messages
<20> leaving system message behind
<21> '''
<22> for h in reversed(history[:-1]):
<23> if h.get("bot"):
<24> messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
<25> messages.insert(1, {"role": self.USER, "content" : h.get("user")</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
token_count = token_count + self.num_tokens_from_messages(messages, "gpt-3.5-turbo")
if token_count > approx_max_tokens*4:
break
return messages
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
at: tiktoken.model
encoding_for_model(model_name: str) -> Encoding
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
+ self.chatgpt_model = chatgpt_model
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
#Chat roles
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
- prompt_prefix = """<|im_start|>system
- Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
- Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
- For tabular information return it as an html table. Do not return markdown format.
- Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
- {follow_up_questions_prompt}
- {injected_prompt}
- Sources:
- {sources}
- <|im_end|>
- {chat_history}
- """
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of Sources:. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use</s>
===========changed ref 2===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# offset: 1
<s>.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
===========changed ref 3===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
use_semantic_captions = True if overrides.get("semantic_captions") else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
# STEP 1: Generate an optimized keyword search query based on the chat history and the last question
prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
completion = openai.Completion.create(
engine=self.gpt_deployment,
prompt=prompt,
temperature=0.0,
max_tokens=32,
n=1,
stop=["\n"])
q = completion.choices[0].text
# STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
if overrides.get("semantic_ranker"):
r = self.search_client.search(q,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
top=top,
query_caption="extractive|highlight-false" if use_semantic_captions else None)
else:
r = self.search_client.search(q, filter=filter, top=top)
if use_semantic_captions:
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else</s>
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
3a0b9c7a8363e38f4c014dabfb09292af703adce
|
add model validation for tiktok
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
if prompt_override is None:
prompt = self.prompt_prefix.format(injected_prompt="", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
prompt = self.prompt_prefix.format(injected_prompt=prompt_override[3:] + "\n", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
else:
prompt = prompt_override.format(sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
print(messages)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature")</s>
===========below chunk 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 2
<s>,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chatContent, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of Sources:. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_messages_from_history(prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 2===========
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.num_tokens_from_messages
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
3a0b9c7a8363e38f4c014dabfb09292af703adce
|
add model validation for tiktok
|
<0>:<add> encoding = tiktoken.encoding_for_model(self.get_oai_chatmodel_tiktok(model))
<del> encoding = tiktoken.encoding_for_model(model)
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def num_tokens_from_messages(self, messages, model: str):
<0> encoding = tiktoken.encoding_for_model(model)
<1> num_tokens = 0
<2> for message in messages:
<3> num_tokens += 4 # every message follows <im_start>{role/name}\n{content}<im_end>\n
<4> for key, value in message.items():
<5> num_tokens += len(encoding.encode(value))
<6> if key == "name": # if there's a name, the role is omitted
<7> num_tokens += -1 # role is always required and always 1 token
<8> num_tokens += 2 # every reply is primed with <im_start>assistant
<9> return num_tokens
<10>
|
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
use_semantic_captions = True if overrides.get("semantic_captions") else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
# STEP 1: Generate an optimized keyword search query based on the chat history and the last question
prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
completion = openai.Completion.create(
engine=self.gpt_deployment,
prompt=prompt,
temperature=0.0,
max_tokens=32,
n=1,
stop=["\n"])
q = completion.choices[0].text
# STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
if overrides.get("semantic_ranker"):
r = self.search_client.search(q,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
top=top,
query_caption="extractive|highlight-false" if use_semantic_captions else None)
else:
r = self.search_client.search(q, filter=filter, top=top)
if use_semantic_captions:
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else</s>
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
<s>ewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
- if prompt_override is None:
- prompt = self.prompt_prefix.format(injected_prompt="", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
- elif prompt_override.startswith(">>>"):
- prompt = self.prompt_prefix.format(injected_prompt=prompt_override[3:] + "\n", sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
- else:
- prompt = prompt_override.format(sources=content, chat_history=self.get_chat_history_as_text(history), follow_up_questions_prompt=follow_up_questions_prompt)
-
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
print(messages)
# STEP 3: Generate a</s>
===========changed ref 2===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 2
<s> content specific answer using the search results and chat history
chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chatContent, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
900eb7811bfdb60529ab969e895ead717ee3ef07
|
s
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
print(messages)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chatContent, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of Sources:. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_messages_from_history(prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 2===========
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_messages_from_history
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
0572986739f72c356bd8e7c749b08f4c51dbae5b
|
address cmments
|
<13>:<add> userContent = history[-1]["user"] + " \nSources:" + sources
<del> userContent = history[-1]["user"] + "\nSources:" + sources
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
<0> messages = []
<1> token_count = 0
<2> if prompt_override is None:
<3> system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
<4> elif prompt_override.startswith(">>>"):
<5> system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
<6> else:
<7> system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
<8>
<9> messages.append({"role":self.SYSTEM, "content": system_message})
<10> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
<11>
<12> #latest conversation
<13> userContent = history[-1]["user"] + "\nSources:" + sources
<14> messages.append({"role": self.USER, "content": userContent})
<15> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
<16>
<17> '''
<18> Enqueue in reverse order
<19> if limit exceeds truncate old messages
<20> leaving system message behind
<21> '''
<22> for h in reversed(history[:-1]):
<23> if h.get("bot"):
<24> messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
<25> messages.insert(1, {"role": self.USER, "content" : h.get("user")})
<26> token_</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
if token_count > approx_max_tokens*4:
break
return messages
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
num_tokens_from_messages(self, messages, model: str)
num_tokens_from_messages(messages, model: str)
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.chatgpt_model = chatgpt_model
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
#Chat roles
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
+ Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
- Answer ONLY with the facts listed in the list of Sources:. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation</s>
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# offset: 1
<s> after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.num_tokens_from_messages
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
0572986739f72c356bd8e7c749b08f4c51dbae5b
|
address cmments
|
<3>:<add> num_tokens += 2 # every message follows {role/name}\n{content}\n
<del> num_tokens += 4 # every message follows <im_start>{role/name}\n{content}<im_end>\n
<8>:<del> num_tokens += 2 # every reply is primed with <im_start>assistant
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def num_tokens_from_messages(self, messages, model: str):
<0> encoding = tiktoken.encoding_for_model(self.get_oai_chatmodel_tiktok(model))
<1> num_tokens = 0
<2> for message in messages:
<3> num_tokens += 4 # every message follows <im_start>{role/name}\n{content}<im_end>\n
<4> for key, value in message.items():
<5> num_tokens += len(encoding.encode(value))
<6> if key == "name": # if there's a name, the role is omitted
<7> num_tokens += -1 # role is always required and always 1 token
<8> num_tokens += 2 # every reply is primed with <im_start>assistant
<9> return num_tokens
<10>
|
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
get_oai_chatmodel_tiktok(aoaimodel: str)
at: tiktoken.core.Encoding
encode(text: str, *, allowed_special: Union[Literal["all"], AbstractSet[str]]=set(), disallowed_special: Union[Literal["all"], Collection[str]]="all") -> list[int]
at: tiktoken.model
encoding_for_model(model_name: str) -> Encoding
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
messages = []
token_count = 0
if prompt_override is None:
system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
else:
system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
messages.append({"role":self.SYSTEM, "content": system_message})
token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
#latest conversation
+ userContent = history[-1]["user"] + " \nSources:" + sources
- userContent = history[-1]["user"] + "\nSources:" + sources
messages.append({"role": self.USER, "content": userContent})
token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
'''
Enqueue in reverse order
if limit exceeds truncate old messages
leaving system message behind
'''
for h in reversed(history[:-1]):
if h.get("bot"):
messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
messages.insert(1, {"role": self.USER, "content" : h.get("user")})
token_count = token_count</s>
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
<s>(1, {"role": self.USER, "content" : h.get("user")})
token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
if token_count > approx_max_tokens*4:
break
return messages
===========changed ref 2===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
#Chat roles
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
+ Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
- Answer ONLY with the facts listed in the list of Sources:. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation</s>
===========changed ref 3===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# offset: 1
<s> after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_messages_from_history
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
5c75cadd2866503945629b07c3ee2f24f7f06a8b
|
Update app/backend/approaches/chatreadretrieveread.py
|
<10>:<add> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<del> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
<0> messages = []
<1> token_count = 0
<2> if prompt_override is None:
<3> system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
<4> elif prompt_override.startswith(">>>"):
<5> system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
<6> else:
<7> system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
<8>
<9> messages.append({"role":self.SYSTEM, "content": system_message})
<10> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
<11>
<12> #latest conversation
<13> userContent = history[-1]["user"] + " \nSources:" + sources
<14> messages.append({"role": self.USER, "content": userContent})
<15> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
<16>
<17> '''
<18> Enqueue in reverse order
<19> if limit exceeds truncate old messages
<20> leaving system message behind
<21> '''
<22> for h in reversed(history[:-1]):
<23> if h.get("bot"):
<24> messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
<25> messages.insert(1, {"role": self.USER, "content" : h.get("user")})
<26> token</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
if token_count > approx_max_tokens*4:
break
return messages
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
num_tokens_from_messages(messages, model: str)
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.chatgpt_model = chatgpt_model
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_messages_from_history
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
929bf01ad8ed14b25a77af9d513a108b9d8df2e3
|
Update app/backend/approaches/chatreadretrieveread.py
|
<15>:<add> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<del> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
<0> messages = []
<1> token_count = 0
<2> if prompt_override is None:
<3> system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
<4> elif prompt_override.startswith(">>>"):
<5> system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
<6> else:
<7> system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
<8>
<9> messages.append({"role":self.SYSTEM, "content": system_message})
<10> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<11>
<12> #latest conversation
<13> userContent = history[-1]["user"] + " \nSources:" + sources
<14> messages.append({"role": self.USER, "content": userContent})
<15> token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
<16>
<17> '''
<18> Enqueue in reverse order
<19> if limit exceeds truncate old messages
<20> leaving system message behind
<21> '''
<22> for h in reversed(history[:-1]):
<23> if h.get("bot"):
<24> messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
<25> messages.insert(1, {"role": self.USER, "content" : h.get("user")})
<26> token_count = token</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
if token_count > approx_max_tokens*4:
break
return messages
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
num_tokens_from_messages(messages, model: str)
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.chatgpt_model = chatgpt_model
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_messages_from_history
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
501c6ca7067479f96b1aea872a954373199d3cfb
|
Update app/backend/approaches/chatreadretrieveread.py
|
<14>:<add> messages.append({"role": self.USER, "content": user_content})
<del> messages.append({"role": self.USER, "content": userContent})
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
<0> messages = []
<1> token_count = 0
<2> if prompt_override is None:
<3> system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
<4> elif prompt_override.startswith(">>>"):
<5> system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
<6> else:
<7> system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
<8>
<9> messages.append({"role":self.SYSTEM, "content": system_message})
<10> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<11>
<12> #latest conversation
<13> userContent = history[-1]["user"] + " \nSources:" + sources
<14> messages.append({"role": self.USER, "content": userContent})
<15> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<16>
<17> '''
<18> Enqueue in reverse order
<19> if limit exceeds truncate old messages
<20> leaving system message behind
<21> '''
<22> for h in reversed(history[:-1]):
<23> if h.get("bot"):
<24> messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
<25> messages.insert(1, {"role": self.USER, "content" : h.get("user")})
<26> token_count = token_count + self</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
if token_count > approx_max_tokens*4:
break
return messages
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
num_tokens_from_messages(messages, model: str)
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.chatgpt_model = chatgpt_model
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
ca9cedb33e9f5bec80dc6d35e377d204ad2b96e6
|
Update app/backend/approaches/chatreadretrieveread.py
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chatContent, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_messages_from_history(prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
===========unchanged ref 2===========
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
2249247fdb3222fa73671027d584e24005515c50
|
fix token estimation and address counts
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chat_content, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_messages_from_history(prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 2===========
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_messages_from_history
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
2249247fdb3222fa73671027d584e24005515c50
|
fix token estimation and address counts
|
<10>:<add> token_count += self.num_tokens_from_messages(messages[-1], self.chatgpt_model)
<del> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<13>:<add> user_content = history[-1]["user"] + " \nSources:" + sources
<del> userContent = history[-1]["user"] + " \nSources:" + sources
<15>:<add> token_count += token_count + self.num_tokens_from_messages(messages[-1], self.chatgpt_model)
<del> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<21>:<add> Keep track of token count for each conversation
<add> If token count exceeds limit, break
<25>:<add> token_count += self.num_tokens_from_messages(messages[1], self.chatgpt_model)
<26>:<add> token_count += self.num_tokens_from_messages(messages[1], self.chatgpt_model)
<del> token_count = token_count +
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
<0> messages = []
<1> token_count = 0
<2> if prompt_override is None:
<3> system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
<4> elif prompt_override.startswith(">>>"):
<5> system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
<6> else:
<7> system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
<8>
<9> messages.append({"role":self.SYSTEM, "content": system_message})
<10> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<11>
<12> #latest conversation
<13> userContent = history[-1]["user"] + " \nSources:" + sources
<14> messages.append({"role": self.USER, "content": user_content})
<15> token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
<16>
<17> '''
<18> Enqueue in reverse order
<19> if limit exceeds truncate old messages
<20> leaving system message behind
<21> '''
<22> for h in reversed(history[:-1]):
<23> if h.get("bot"):
<24> messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
<25> messages.insert(1, {"role": self.USER, "content" : h.get("user")})
<26> token_count = token_count +</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
if token_count > approx_max_tokens*4:
break
return messages
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
num_tokens_from_messages(messages, model: str)
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.chatgpt_model = chatgpt_model
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
use_semantic_captions = True if overrides.get("semantic_captions") else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
# STEP 1: Generate an optimized keyword search query based on the chat history and the last question
prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
completion = openai.Completion.create(
engine=self.gpt_deployment,
prompt=prompt,
temperature=0.0,
max_tokens=32,
n=1,
stop=["\n"])
q = completion.choices[0].text
# STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
if overrides.get("semantic_ranker"):
r = self.search_client.search(q,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
top=top,
query_caption="extractive|highlight-false" if use_semantic_captions else None)
else:
r = self.search_client.search(q, filter=filter, top=top)
if use_semantic_captions:
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else</s>
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
<s>ewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
+ chat_completion = openai.ChatCompletion.create(
- chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
+ chat_content = chat_completion.choices[0].message.content
- chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chat_content, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.num_tokens_from_messages
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
2249247fdb3222fa73671027d584e24005515c50
|
fix token estimation and address counts
|
<2>:<del> for message in messages:
<3>:<add> num_tokens += 2 # every message follows {role/name}\n{content}\n
<del> num_tokens += 2 # every message follows {role/name}\n{content}\n
<4>:<add> for key, value in message.items():
<del> for key, value in message.items():
<5>:<add> num_tokens += len(encoding.encode(value))
<del> num_tokens += len(encoding.encode(value))
<6>:<add> if key == "name": # if there's a name, the role is omitted
<del> if key == "name": # if there's a name, the role is omitted
<7>:<add> num_tokens += -1 # role is always required and always 1 token
<del> num_tokens += -1 # role is always required and always 1 token
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
-
+ def num_tokens_from_messages(self, message: any, model: str):
- def num_tokens_from_messages(self, messages, model: str):
<0> encoding = tiktoken.encoding_for_model(self.get_oai_chatmodel_tiktok(model))
<1> num_tokens = 0
<2> for message in messages:
<3> num_tokens += 2 # every message follows {role/name}\n{content}\n
<4> for key, value in message.items():
<5> num_tokens += len(encoding.encode(value))
<6> if key == "name": # if there's a name, the role is omitted
<7> num_tokens += -1 # role is always required and always 1 token
<8> return num_tokens
<9>
|
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.get_messages_from_history
messages = []
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
messages = []
token_count = 0
if prompt_override is None:
system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
else:
system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
messages.append({"role":self.SYSTEM, "content": system_message})
+ token_count += self.num_tokens_from_messages(messages[-1], self.chatgpt_model)
- token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
#latest conversation
+ user_content = history[-1]["user"] + " \nSources:" + sources
- userContent = history[-1]["user"] + " \nSources:" + sources
messages.append({"role": self.USER, "content": user_content})
+ token_count += token_count + self.num_tokens_from_messages(messages[-1], self.chatgpt_model)
- token_count += self.num_tokens_from_messages(messages, self.chatgpt_model)
'''
Enqueue in reverse order
if limit exceeds truncate old messages
leaving system message behind
+ Keep track of token count for each conversation
+ If token count exceeds limit, break
'''
for h in reversed</s>
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Generate messages needed for chat Completion api
def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
<s>
+ Keep track of token count for each conversation
+ If token count exceeds limit, break
'''
for h in reversed(history[:-1]):
if h.get("bot"):
messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
+ token_count += self.num_tokens_from_messages(messages[1], self.chatgpt_model)
messages.insert(1, {"role": self.USER, "content" : h.get("user")})
+ token_count += self.num_tokens_from_messages(messages[1], self.chatgpt_model)
- token_count = token_count + self.num_tokens_from_messages(messages, self.chatgpt_model)
if token_count > approx_max_tokens*4:
break
-
return messages
===========changed ref 2===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
use_semantic_captions = True if overrides.get("semantic_captions") else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
# STEP 1: Generate an optimized keyword search query based on the chat history and the last question
prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
completion = openai.Completion.create(
engine=self.gpt_deployment,
prompt=prompt,
temperature=0.0,
max_tokens=32,
n=1,
stop=["\n"])
q = completion.choices[0].text
# STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
if overrides.get("semantic_ranker"):
r = self.search_client.search(q,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
top=top,
query_caption="extractive|highlight-false" if use_semantic_captions else None)
else:
r = self.search_client.search(q, filter=filter, top=top)
if use_semantic_captions:
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else</s>
===========changed ref 3===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
<s>ewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
+ chat_completion = openai.ChatCompletion.create(
- chatCompletion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
+ chat_content = chat_completion.choices[0].message.content
- chatContent = chatCompletion.choices[0].message.content
return {"data_points": results, "answer": chat_content, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.num_tokens_from_messages
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
81239a79d2eb392a78778e90b2a23269082fc58d
|
address comments'
|
<0>:<add> """
<add> Calculate the number of tokens required to encode a message.
<add> Args:
<add> message (any): The message to encode, represented as a dictionary.
<add> model (str): The name of the model to use for encoding.
<add> Returns:
<add> int: The total number of tokens required to encode the message.
<add> Example:
<add> message = {'role': 'user', 'name': 'John', 'content': 'Hello, how are you?'}
<add> model = 'gpt-3.5-turbo'
<add> num_tokens_from_messages(message, model)
<add> output: 11
<add> """
<5>:<del> if key == "name": # if there's a name, the role is omitted
<6>:<del> num_tokens += -1 # role is always required and always 1 token
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+
+ def num_tokens_from_messages(self, message: dict[str,str], model: str):
- def num_tokens_from_messages(self, message: any, model: str):
<0> encoding = tiktoken.encoding_for_model(self.get_oai_chatmodel_tiktok(model))
<1> num_tokens = 0
<2> num_tokens += 2 # every message follows {role/name}\n{content}\n
<3> for key, value in message.items():
<4> num_tokens += len(encoding.encode(value))
<5> if key == "name": # if there's a name, the role is omitted
<6> num_tokens += -1 # role is always required and always 1 token
<7> return num_tokens
<8>
|
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.num_tokens_from_messages
encoding = tiktoken.encoding_for_model(self.get_oai_chatmodel_tiktok(model))
at: tiktoken.core.Encoding
encode(text: str, *, allowed_special: Union[Literal["all"], AbstractSet[str]]=set(), disallowed_special: Union[Literal["all"], Collection[str]]="all") -> list[int]
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
-
- '''
- Source: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb
- Adapted: https://learn.microsoft.com/en-us/azure/cognitive-services/openai/how-to/chatgpt?pivots=programming-language-chat-completions#managing-conversations
-
- Method takes in a single conversation and calculate prompt tokens
- for chat api
-
- Keys role and content are accounted seperately.
-
- Values of content are encoded by model type and calculated the length.
-
- This gives close proximity of token length measurement used in gpt models
-
- message = {"role":"assistant", "content":"how can I assist you?"}
- '''
-
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
e3dc08714d976fe1f112b5e9fdcb758dda9d619f
|
Merge pull request #407 from Azure-Samples/srbalakr/update-nb
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chat_completion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chat_content = chat_completion.choices[0].message.content
return {"data_points": results, "answer": chat_content, "thoughts": f"Searched for:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
If the question is not in English, translate the question to English before generating the search query.
Chat History:
{chat_history}
Question:
{question}
Search query:
"""
get_chat_history_as_text(history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str
get_messages_from_history(prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
===========unchanged ref 2===========
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
|
scripts.prepdocs/create_sections
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
755188cf7c41e0909904e3d81a001e39640ffe0b
|
Handle non-ascii filenames (#418)
|
<0>:<add> file_id = filename_to_id(filename)
<2>:<add> "id": f"{file_id}-page-{i}",
<del> "id": re.sub("[^0-9a-zA-Z_-]","_",f"{filename}-{i}"),
|
# module: scripts.prepdocs
def create_sections(filename, page_map):
<0> for i, (section, pagenum) in enumerate(split_text(page_map)):
<1> yield {
<2> "id": re.sub("[^0-9a-zA-Z_-]","_",f"{filename}-{i}"),
<3> "content": section,
<4> "category": args.category,
<5> "sourcepage": blob_name_from_file_page(filename, pagenum),
<6> "sourcefile": filename
<7> }
<8>
|
===========unchanged ref 0===========
at: scripts.prepdocs
args = parser.parse_args()
at: scripts.prepdocs.create_search_index
index_client = SearchIndexClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
credential=search_creds)
index = SearchIndex(
name=args.index,
fields=[
SimpleField(name="id", type="Edm.String", key=True),
SearchableField(name="content", type="Edm.String", analyzer_name="en.microsoft"),
SimpleField(name="category", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcepage", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcefile", type="Edm.String", filterable=True, facetable=True)
],
semantic_settings=SemanticSettings(
configurations=[SemanticConfiguration(
name='default',
prioritized_fields=PrioritizedFields(
title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))])
)
===========changed ref 0===========
# module: scripts.prepdocs
+ def filename_to_id(filename):
+ filename_ascii = re.sub("[^0-9a-zA-Z_-]", "_", filename)
+ filename_hash = base64.b16encode(filename.encode('utf-8')).decode('ascii')
+ return f"file-{filename_ascii}-{filename_hash}"
+
===========changed ref 1===========
+ # module: scripts.test_prepdocs
+
+
===========changed ref 2===========
+ # module: scripts.test_prepdocs
+ def test_filename_to_id():
+ # test ascii filename
+ assert filename_to_id("foo.pdf") == "file-foo_pdf-666F6F2E706466"
+ # test filename containing unicode
+ assert filename_to_id("foo\u00A9.txt") == "file-foo__txt-666F6FC2A92E747874"
+ # test filenaming starting with unicode
+ assert filename_to_id("ファイル名.pdf") == "file-______pdf-E38395E382A1E382A4E383ABE5908D2E706466"
+
===========changed ref 3===========
# module: scripts.prepdocs
MAX_SECTION_LENGTH = 1000
SENTENCE_SEARCH_LIMIT = 100
SECTION_OVERLAP = 100
- parser = argparse.ArgumentParser(
- description="Prepare documents by extracting content from PDFs, splitting content into sections, uploading to blob storage, and indexing in a search index.",
- epilog="Example: prepdocs.py '..\data\*' --storageaccount myaccount --container mycontainer --searchservice mysearch --index myindex -v"
- )
- parser.add_argument("files", help="Files to be processed")
- parser.add_argument("--category", help="Value for the category field in the search index for all sections indexed in this run")
- parser.add_argument("--skipblobs", action="store_true", help="Skip uploading individual pages to Azure Blob Storage")
- parser.add_argument("--storageaccount", help="Azure Blob Storage account name")
- parser.add_argument("--container", help="Azure Blob Storage container name")
- parser.add_argument("--storagekey", required=False, help="Optional. Use this Azure Blob Storage account key instead of the current user identity to login (use az login to set current user for Azure)")
- parser.add_argument("--tenantid", required=False, help="Optional. Use this to define the Azure directory where to authenticate)")
- parser.add_argument("--searchservice", help="Name of the Azure Cognitive Search service where content should be indexed (must exist already)")
- parser.add_argument("--index", help="Name of the Azure Cognitive Search index where content should be indexed (will be created if it doesn't exist)")
- parser.add_argument("--searchkey", required=False, help="Optional. Use this Azure Cognitive Search account key instead of the current user identity to login (use az login to set current user for Azure)")
- parser.add_argument("--remove", action="store_true", help="Remove references to this document from blob storage and the search index")
- parser.add_argument("--removeall", action="store_true", help="Remove all blobs from blob storage and documents from the search index")
- parser.add_argument("--localpdfparser</s>
===========changed ref 4===========
# module: scripts.prepdocs
# offset: 1
<s>_true", help="Remove all blobs from blob storage and documents from the search index")
- parser.add_argument("--localpdfparser", action="store_true", help="Use PyPdf local PDF parser (supports only digital PDFs) instead of Azure Form Recognizer service to extract text, tables and layout from the documents")
- parser.add_argument("--formrecognizerservice", required=False, help="Optional. Name of the Azure Form Recognizer service which will be used to extract text, tables and layout from the documents (must exist already)")
- parser.add_argument("--formrecognizerkey", required=False, help="Optional. Use this Azure Form Recognizer account key instead of the current user identity to login (use az login to set current user for Azure)")
- parser.add_argument("--verbose", "-v", action="store_true", help="Verbose output")
- args = parser.parse_args()
+ if __name__ == "__main__":
- # Use the current user identity to connect to Azure services unless a key is explicitly set for any of them
- azd_credential = AzureDeveloperCliCredential() if args.tenantid == None else AzureDeveloperCliCredential(tenant_id=args.tenantid, process_timeout=60)
- default_creds = azd_credential if args.searchkey == None or args.storagekey == None else None
- search_creds = default_creds if args.searchkey == None else AzureKeyCredential(args.searchkey)
- if not args.skipblobs:
- storage_creds = default_creds if args.storagekey == None else args.storagekey
- if not args.localpdfparser:
- # check if Azure Form Recognizer credentials are provided
- if args.formrecognizerservice == None:
- print("Error: Azure Form Recognizer service is not provided. Please provide formrecognizerservice or use --localpdfparser for local pypdf parser.")
+ parser = argparse.ArgumentParser(
+ description="Prepare documents by</s>
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6bfb2ccfdb0c2e15a3285465b031939a43f11eac
|
Migration Completion api to chat completion api (#419)
|
<3>:<del> self.gpt_deployment = gpt_deployment
<6>:<add> self.chatgpt_token_limit = get_token_limit(chatgpt_model)
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
<0> self.search_client = search_client
<1> self.chatgpt_deployment = chatgpt_deployment
<2> self.chatgpt_model = chatgpt_model
<3> self.gpt_deployment = gpt_deployment
<4> self.sourcepage_field = sourcepage_field
<5> self.content_field = content_field
<6>
|
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
Do not include any special characters like '+'.
If the question is not in English, translate the question to English before generating the search query.
Search Query:
"""
query_prompt_few_shots = [
{'role' : USER, 'content' : 'What are my health plans?' },
{'role' : ASSISTANT, 'content' : 'Show available health plans' },
{'role' : USER, 'content' : 'does my plan cover cardio?' },
{'role' : ASSISTANT, 'content' : 'Health plan cardio coverage' }
]
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Chat roles
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include c</s>
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# offset: 1
<s> plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
+ Do not include any special characters like '+'.
If the question is not in English, translate the question to English before generating the search query.
+ Search Query:
+ """
+ query_prompt_few_shots = [
+ {'role' : USER, 'content' : 'What are my health plans?' },
+ {'role' : ASSISTANT, 'content' : 'Show available health plans' },
+ {'role' : USER, 'content' : 'does my plan cover cardio?' },
+ {'role' : ASSISTANT, 'content' : 'Health plan cardio coverage' }
+ ]
- Chat History:
- {chat_history}
- Question:
- {question}
-
- Search query:
- """
-
===========changed ref 2===========
+ # module: app.backend.core.messagebuilder
+
+
===========changed ref 3===========
+ # module: app.backend.core.modelhelper
+
+
===========changed ref 4===========
+ # module: app.backend.core.messagebuilder
+ class MessageBuilder:
+ def append_message(self, role: str, content: str, index: int = 1):
+ self.messages.insert(index, {'role': role, 'content': content})
+ self.token_length += num_tokens_from_messages(
+ self.messages[index], self.model)
+
===========changed ref 5===========
+ # module: app.backend.core.modelhelper
+ def get_oai_chatmodel_tiktok(aoaimodel: str) -> str:
+ if aoaimodel == "" or aoaimodel is None:
+ raise ValueError("Expected AOAI chatGPT model name")
+
+ return AOAI_2_OAI.get(aoaimodel)
+
===========changed ref 6===========
+ # module: app.backend.core.modelhelper
+ def get_token_limit(model_id: str) -> int:
+ if model_id not in MODELS_2_TOKEN_LIMITS:
+ raise ValueError("Expected Model Gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS.get(model_id)
+
===========changed ref 7===========
+ # module: app.backend.core.messagebuilder
+ class MessageBuilder:
+ def __init__(self, system_content: str, chatgpt_model: str):
+ self.messages = [{'role': 'system', 'content': system_content}]
+ self.model = chatgpt_model
+ self.token_length = num_tokens_from_messages(
+ self.messages[-1], self.model)
+
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6bfb2ccfdb0c2e15a3285465b031939a43f11eac
|
Migration Completion api to chat completion api (#419)
|
<5>:<add> user_q = 'Generate search query for: ' + history[-1]["user"]
<add>
<6>:<add> messages = self.get_messages_from_history(
<add> self.query_prompt_template,
<add> self.chatgpt_model,
<add> history,
<add> user_q,
<add> self.query_prompt_few_shots,
<add> self.chatgpt_token_limit - len(user_q)
<add> )
<add>
<del> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7>:<add> chat_completion = openai.ChatCompletion.create(
<del> completion = openai.Completion.create(
<8>:<add> deployment_id=self.chatgpt_deployment,
<del> engine=self.gpt_deployment,
<9>:<add> model=self.chatgpt_model,
<add> messages=messages,
<del> prompt=prompt,
<12>:<add> n=1)
<del> n=1,
<13>:<add>
<del> stop=["\n"])
<14>:<add> q = chat_completion.choices[0].message.content
<del> q = completion.choices[0].text
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<6> prompt = self.query_prompt_template.format(chat_history=self.get_chat_history_as_text(history, include_last_turn=False), question=history[-1]["user"])
<7> completion = openai.Completion.create(
<8> engine=self.gpt_deployment,
<9> prompt=prompt,
<10> temperature=0.0,
<11> max_tokens=32,
<12> n=1,
<13> stop=["\n"])
<14> q = completion.choices[0].text
<15>
<16> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<17> if overrides.get("semantic_ranker"):
<18> r = self.search_client.search(q,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<26> else:
<27> r = self.search_client.search(q, filter=filter, top=top)
<28> if use_semantic_captions:
<29> results = [doc[self.sourcepage_field] + ": " + non</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_template")
messages = self.get_messages_from_history(prompt_override=prompt_override, follow_up_questions_prompt=follow_up_questions_prompt,history=history, sources=content)
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
chat_completion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chat_content = chat_completion.choices[0].message.content
msg_to_display = '\n\n'.join([str(message) for message in messages])
return {"data_points": results, "answer": chat_content, "thoughts": f"Searched for:<br>{q}<br><br>Conversations:<br>" + msg_to_display.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
Do not include any special characters like '+'.
If the question is not in English, translate the question to English before generating the search query.
Search Query:
"""
query_prompt_few_shots = [
{'role' : USER, 'content' : 'What are my health plans?' },
{'role' : ASSISTANT, 'content' : 'Show available health plans' },
{'role' : USER, 'content' : 'does my plan cover cardio?' },
{'role' : ASSISTANT, 'content' : 'Health plan cardio coverage' }
]
===========unchanged ref 1===========
get_messages_from_history(system_prompt: str, model_id: str, history: Sequence[dict[str, str]], user_conv: str, few_shots=[], max_tokens: int=4096) -> []
get_messages_from_history(self, system_prompt: str, model_id: str, history: Sequence[dict[str, str]], user_conv: str, few_shots=[], max_tokens: int=4096) -> []
get_messages_from_history(prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int=1000) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: core.modelhelper
get_token_limit(model_id: str) -> int
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
===========unchanged ref 2===========
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
- self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
+ self.chatgpt_token_limit = get_token_limit(chatgpt_model)
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_messages_from_history
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6bfb2ccfdb0c2e15a3285465b031939a43f11eac
|
Migration Completion api to chat completion api (#419)
|
<0>:<del> '''
<1>:<del> Generate messages needed for chat Completion api
<2>:<del> '''
<3>:<del> messages = []
<4>:<del> token_count = 0
<5>:<del> if prompt_override is None:
<6>:<del> system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
<7>:<del> elif prompt_override.startswith(">>>"):
<8>:<del> system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
<9>:<del> else:
<10>:<del> system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
<11>:<add> message_builder = MessageBuilder(system_prompt, model_id)
<12>:<del> messages.append({"role":self.SYSTEM, "content": system_message})
<13>:<del> token_count += self.num_tokens_from_messages(messages[-1], self.chatgpt_model)
<14>:<del>
<15>:<del> # latest conversation
<16>:<del> user_content = history[-1]["user"] + " \nSources:" + sources
<17>:<del> messages.append({"role": self.USER, "content": user_content})
<18>:<del> token_count += token_count + self.num_tokens_from_messages(messages[-1], self.chatgpt_model)
<19>:<add> # Add examples to show the chat what responses we want. It will try to mimic any responses and make sure they match the rules laid out in the system message.
<add> for shot in few_shots:
<add> message_builder.append_message(shot.get('role'), shot.get('content'))
<20>:<del> '''
<21>:<del> Enqueue in reverse order
<22>:<del> if limit exceeds truncate old messages
<23>:<del> leaving system message behind
<24>:<del> Keep track of token count for each conversation
<25>:<del> If token count exceeds limit, break
<26>:<del> '''
<27>:<add> user_content = user_conv
<add> append_index = len(few_shots) + 1
<add>
<add> message_builder.append_message(self.USER, user_content, index=append_index)
<add>
|
<s>ReadRetrieveReadApproach(Approach):
+ def get_messages_from_history(self, system_prompt: str, model_id: str, history: Sequence[dict[str, str]], user_conv: str, few_shots = [], max_tokens: int = 4096) -> []:
- def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
<0> '''
<1> Generate messages needed for chat Completion api
<2> '''
<3> messages = []
<4> token_count = 0
<5> if prompt_override is None:
<6> system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
<7> elif prompt_override.startswith(">>>"):
<8> system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
<9> else:
<10> system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
<11>
<12> messages.append({"role":self.SYSTEM, "content": system_message})
<13> token_count += self.num_tokens_from_messages(messages[-1], self.chatgpt_model)
<14>
<15> # latest conversation
<16> user_content = history[-1]["user"] + " \nSources:" + sources
<17> messages.append({"role": self.USER, "content": user_content})
<18> token_count += token_count + self.num_tokens_from_messages(messages[-1], self.chatgpt_model)
<19>
<20> '''
<21> Enqueue in reverse order
<22> if limit exceeds truncate old messages
<23> leaving system message behind
<24> Keep track of token count for each conversation
<25> If token count exceeds limit, break
<26> '''
<27> for h in reversed(history[:</s>
|
===========below chunk 0===========
<s>(Approach):
+ def get_messages_from_history(self, system_prompt: str, model_id: str, history: Sequence[dict[str, str]], user_conv: str, few_shots = [], max_tokens: int = 4096) -> []:
- def get_messages_from_history(self, prompt_override, follow_up_questions_prompt, history: Sequence[dict[str, str]], sources: str, approx_max_tokens: int = 1000) -> []:
# offset: 1
if h.get("bot"):
messages.insert(1, {"role": self.ASSISTANT, "content" : h.get("bot")})
token_count += self.num_tokens_from_messages(messages[1], self.chatgpt_model)
messages.insert(1, {"role": self.USER, "content" : h.get("user")})
token_count += self.num_tokens_from_messages(messages[1], self.chatgpt_model)
if token_count > approx_max_tokens*4:
break
return messages
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.run
content = "\n".join(results)
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
-
- def get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str:
- history_text = ""
- for h in reversed(history if include_last_turn else history[:-1]):
- history_text = """<|im_start|>user""" + "\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot", "") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
- if len(history_text) > approx_max_tokens*4:
- break
- return history_text
-
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
- self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
+ self.chatgpt_token_limit = get_token_limit(chatgpt_model)
===========changed ref 2===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Chat roles
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include c</s>
===========changed ref 3===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# offset: 1
<s> plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
+ Do not include any special characters like '+'.
If the question is not in English, translate the question to English before generating the search query.
+ Search Query:
+ """
+ query_prompt_few_shots = [
+ {'role' : USER, 'content' : 'What are my health plans?' },
+ {'role' : ASSISTANT, 'content' : 'Show available health plans' },
+ {'role' : USER, 'content' : 'does my plan cover cardio?' },
+ {'role' : ASSISTANT, 'content' : 'Health plan cardio coverage' }
+ ]
- Chat History:
- {chat_history}
- Question:
- {question}
-
- Search query:
- """
-
===========changed ref 4===========
+ # module: app.backend.core.messagebuilder
+
+
===========changed ref 5===========
+ # module: app.backend.core.modelhelper
+
+
|
app.backend.approaches.retrievethenread/RetrieveThenReadApproach.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6bfb2ccfdb0c2e15a3285465b031939a43f11eac
|
Migration Completion api to chat completion api (#419)
|
<2>:<add> self.chatgpt_model = chatgpt_model
|
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
<0> self.search_client = search_client
<1> self.openai_deployment = openai_deployment
<2> self.sourcepage_field = sourcepage_field
<3> self.content_field = content_field
<4>
|
===========changed ref 0===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
+ system_chat_template = \
- template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
+ "If you cannot answer using the sources below, say you don't know. Use below example to answer"
- "If you cannot answer using the sources below, say you don't know. " + \
- """
+ #shots/sample conversation
+ question = """
- ###
+ 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
- Question: 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
+ """
+ </s>
===========changed ref 1===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
# offset: 1
<s>
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
+ """
+ answer = "In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf]."
- Answer:
- In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf].
-
- ###
- Question: '{q}'?
-
- Sources:
- {retrieved}
-
- Answer:
- """
-
===========changed ref 2===========
+ # module: app.backend.core.messagebuilder
+
+
===========changed ref 3===========
+ # module: app.backend.core.modelhelper
+
+
===========changed ref 4===========
+ # module: app.backend.core.messagebuilder
+ class MessageBuilder:
+ def append_message(self, role: str, content: str, index: int = 1):
+ self.messages.insert(index, {'role': role, 'content': content})
+ self.token_length += num_tokens_from_messages(
+ self.messages[index], self.model)
+
===========changed ref 5===========
+ # module: app.backend.core.modelhelper
+ def get_oai_chatmodel_tiktok(aoaimodel: str) -> str:
+ if aoaimodel == "" or aoaimodel is None:
+ raise ValueError("Expected AOAI chatGPT model name")
+
+ return AOAI_2_OAI.get(aoaimodel)
+
===========changed ref 6===========
+ # module: app.backend.core.modelhelper
+ def get_token_limit(model_id: str) -> int:
+ if model_id not in MODELS_2_TOKEN_LIMITS:
+ raise ValueError("Expected Model Gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS.get(model_id)
+
===========changed ref 7===========
+ # module: app.backend.core.messagebuilder
+ class MessageBuilder:
+ def __init__(self, system_content: str, chatgpt_model: str):
+ self.messages = [{'role': 'system', 'content': system_content}]
+ self.model = chatgpt_model
+ self.token_length = num_tokens_from_messages(
+ self.messages[-1], self.model)
+
===========changed ref 8===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
- def get_oai_chatmodel_tiktok(self, aoaimodel: str):
- if aoaimodel == "" or aoaimodel is None:
- raise Exception("Expected AOAI chatGPT model name")
-
- return "gpt-3.5-turbo" if aoaimodel == "gpt-35-turbo" else aoaimodel
-
===========changed ref 9===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, gpt_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
- self.gpt_deployment = gpt_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
+ self.chatgpt_token_limit = get_token_limit(chatgpt_model)
===========changed ref 10===========
+ # module: app.backend.core.messagebuilder
+ class MessageBuilder:
+ """
+ A class for building and managing messages in a chat conversation.
+ Attributes:
+ message (list): A list of dictionaries representing chat messages.
+ model (str): The name of the ChatGPT model.
+ token_count (int): The total number of tokens in the conversation.
+ Methods:
+ __init__(self, system_content: str, chatgpt_model: str): Initializes the MessageBuilder instance.
+ append_message(self, role: str, content: str, index: int = 1): Appends a new message to the conversation.
+ """
+
===========changed ref 11===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
-
- def get_chat_history_as_text(self, history: Sequence[dict[str, str]], include_last_turn: bool=True, approx_max_tokens: int=1000) -> str:
- history_text = ""
- for h in reversed(history if include_last_turn else history[:-1]):
- history_text = """<|im_start|>user""" + "\n" + h["user"] + "\n" + """<|im_end|>""" + "\n" + """<|im_start|>assistant""" + "\n" + (h.get("bot", "") + """<|im_end|>""" if h.get("bot") else "") + "\n" + history_text
- if len(history_text) > approx_max_tokens*4:
- break
- return history_text
-
|
app.backend.approaches.retrievethenread/RetrieveThenReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
6bfb2ccfdb0c2e15a3285465b031939a43f11eac
|
Migration Completion api to chat completion api (#419)
|
<22>:<add> message_builder = MessageBuilder(overrides.get("prompt_template") or self.system_chat_template, self.chatgpt_model);
<add>
<add> # add user question
<add> user_content = q + "\n" + "Sources:\n {content}".format(content=content)
<add> message_builder.append_message('user', user_content)
<add>
<add> # Add shots/samples. This helps model to mimic response and make sure they match rules laid out in system message.
<add> message_builder.append_message('assistant', self.answer)
<add> message_builder.append_message('user', self.question)
<add>
<add> messages = message_builder.messages
<del> prompt = (overrides.get("prompt_template") or self.template).format(q=q, retrieved=content)
<23>:<add> chat_completion = openai.ChatCompletion.create(
<del> completion = openai.Completion.create(
<24>:<add> deployment_id=self.openai_deployment,
<del> engine=self.openai_deployment,
<25>:<add> model=self.chatgpt_model,
<add> messages=messages,
<del> prompt=prompt,
<28>:<add> n=1
|
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
def run(self, q: str, overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> if overrides.get("semantic_ranker"):
<6> r = self.search_client.search(q,
<7> filter=filter,
<8> query_type=QueryType.SEMANTIC,
<9> query_language="en-us",
<10> query_speller="lexicon",
<11> semantic_configuration_name="default",
<12> top=top,
<13> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<14> else:
<15> r = self.search_client.search(q, filter=filter, top=top)
<16> if use_semantic_captions:
<17> results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
<18> else:
<19> results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
<20> content = "\n".join(results)
<21>
<22> prompt = (overrides.get("prompt_template") or self.template).format(q=q, retrieved=content)
<23> completion = openai.Completion.create(
<24> engine=self.openai_deployment,
<25> prompt=prompt,
<26> temperature=overrides.get("temperature") or 0.3,
<27> max_tokens=1024,
<28> n=1,
<29> stop=["</s>
|
===========below chunk 0===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
return {"data_points": results, "answer": completion.choices[0].text, "thoughts": f"Question:<br>{q}<br><br>Prompt:<br>" + prompt.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.retrievethenread.RetrieveThenReadApproach
template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
"If you cannot answer using the sources below, say you don't know. " + \
"""
###
Question: 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
Answer:
In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf].
###
Question: '{q}'?
Sources:
{retrieved}
Answer:
"""
at: app.backend.approaches.retrievethenread.RetrieveThenReadApproach.__init__
self.search_client = search_client
self.openai_deployment = openai_deployment
===========unchanged ref 1===========
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: openai.api_resources.completion
Completion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.completion.Completion
OBJECT_NAME = "completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
+ self.chatgpt_model = chatgpt_model
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 1===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
+ system_chat_template = \
- template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
+ "If you cannot answer using the sources below, say you don't know. Use below example to answer"
- "If you cannot answer using the sources below, say you don't know. " + \
- """
+ #shots/sample conversation
+ question = """
- ###
+ 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
- Question: 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
+ """
+ </s>
===========changed ref 2===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
# offset: 1
<s>
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
+ """
+ answer = "In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf]."
- Answer:
- In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf].
-
- ###
- Question: '{q}'?
-
- Sources:
- {retrieved}
-
- Answer:
- """
-
===========changed ref 3===========
+ # module: app.backend.core.messagebuilder
+
+
===========changed ref 4===========
+ # module: app.backend.core.modelhelper
+
+
===========changed ref 5===========
+ # module: app.backend.core.messagebuilder
+ class MessageBuilder:
+ def append_message(self, role: str, content: str, index: int = 1):
+ self.messages.insert(index, {'role': role, 'content': content})
+ self.token_length += num_tokens_from_messages(
+ self.messages[index], self.model)
+
|
scripts.prepdocs/create_sections
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<1>:<add> for i, (content, pagenum) in enumerate(split_text(page_map)):
<del> for i, (section, pagenum) in enumerate(split_text(page_map)):
<2>:<add> section = {
<del> yield {
<4>:<add> "content": content,
<del> "content": section,
<9>:<add> if use_vectors:
<add> section["embedding"] = compute_embedding(content)
<add> yield section
|
# module: scripts.prepdocs
+ def create_sections(filename, page_map, use_vectors):
- def create_sections(filename, page_map):
<0> file_id = filename_to_id(filename)
<1> for i, (section, pagenum) in enumerate(split_text(page_map)):
<2> yield {
<3> "id": f"{file_id}-page-{i}",
<4> "content": section,
<5> "category": args.category,
<6> "sourcepage": blob_name_from_file_page(filename, pagenum),
<7> "sourcefile": filename
<8> }
<9>
|
===========unchanged ref 0===========
at: scripts.prepdocs
blob_name_from_file_page(filename, page=0)
split_text(page_map)
filename_to_id(filename)
args = parser.parse_args()
|
scripts.prepdocs/create_search_index
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<9>:<add> SearchField(name="embedding", type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
<add> hidden=False, searchable=True, filterable=False, sortable=False, facetable=False,
<add> vector_search_dimensions=1536, vector_search_configuration="default"),
<17>:<add> title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))]),
<del> title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))])
<18>:<add> vector_search=VectorSearch(
<add> algorithm_configurations=[
<add> VectorSearchAlgorithmConfiguration(
<add> name="default",
<add> kind="hnsw",
<add> hnsw_parameters=HnswParameters(metric="cosine")
<add> )
<add> ]
<add> )
<add> )
<del> )
|
# module: scripts.prepdocs
def create_search_index():
<0> if args.verbose: print(f"Ensuring search index {args.index} exists")
<1> index_client = SearchIndexClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
<2> credential=search_creds)
<3> if args.index not in index_client.list_index_names():
<4> index = SearchIndex(
<5> name=args.index,
<6> fields=[
<7> SimpleField(name="id", type="Edm.String", key=True),
<8> SearchableField(name="content", type="Edm.String", analyzer_name="en.microsoft"),
<9> SimpleField(name="category", type="Edm.String", filterable=True, facetable=True),
<10> SimpleField(name="sourcepage", type="Edm.String", filterable=True, facetable=True),
<11> SimpleField(name="sourcefile", type="Edm.String", filterable=True, facetable=True)
<12> ],
<13> semantic_settings=SemanticSettings(
<14> configurations=[SemanticConfiguration(
<15> name='default',
<16> prioritized_fields=PrioritizedFields(
<17> title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))])
<18> )
<19> if args.verbose: print(f"Creating {args.index} search index")
<20> index_client.create_index(index)
<21> else:
<22> if args.verbose: print(f"Search index {args.index} already exists")
<23>
|
===========unchanged ref 0===========
at: openai.api_resources.embedding
Embedding(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.embedding.Embedding
OBJECT_NAME = "embeddings"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: scripts.prepdocs
args = parser.parse_args()
search_creds = default_creds if args.searchkey == None else AzureKeyCredential(args.searchkey)
at: tenacity
retry(stop: "StopBaseT"=stop_never, wait: "WaitBaseT"=wait_none(), retry: "RetryBaseT"=retry_if_exception_type(), before: t.Callable[["RetryCallState"], None]=before_nothing, after: t.Callable[["RetryCallState"], None]=after_nothing, before_sleep: t.Optional[t.Callable[["RetryCallState"], None]]=None, reraise: bool=False, retry_error_cls: t.Type[RetryError]=RetryError, retry_error_callback: t.Optional[t.Callable[["RetryCallState"], t.Any]]=None, *, sleep: t.Callable[[t.Union[int, float]], None]=sleep) -> t.Any
at: tenacity.stop
stop_after_attempt(max_attempt_number: int)
===========unchanged ref 1===========
at: tenacity.wait
wait_random_exponential(multiplier: typing.Union[int, float]=1, max: _utils.time_unit_type=_utils.MAX_WAIT, exp_base: typing.Union[int, float]=2, min: _utils.time_unit_type=0)
===========changed ref 0===========
# module: scripts.prepdocs
+ def create_sections(filename, page_map, use_vectors):
- def create_sections(filename, page_map):
file_id = filename_to_id(filename)
+ for i, (content, pagenum) in enumerate(split_text(page_map)):
- for i, (section, pagenum) in enumerate(split_text(page_map)):
+ section = {
- yield {
"id": f"{file_id}-page-{i}",
+ "content": content,
- "content": section,
"category": args.category,
"sourcepage": blob_name_from_file_page(filename, pagenum),
"sourcefile": filename
}
+ if use_vectors:
+ section["embedding"] = compute_embedding(content)
+ yield section
|
app.backend.approaches.readretrieveread/ReadRetrieveReadApproach.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<2>:<add> self.embedding_deployment = embedding_deployment
|
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
<0> self.search_client = search_client
<1> self.openai_deployment = openai_deployment
<2> self.sourcepage_field = sourcepage_field
<3> self.content_field = content_field
<4>
|
===========changed ref 0===========
# module: scripts.prepdocs
+ def before_retry_sleep(retry_state):
+ if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
+
===========changed ref 1===========
# module: scripts.prepdocs
+ @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(15), before_sleep=before_retry_sleep)
+ def compute_embedding(text):
+ return openai.Embedding.create(engine=args.openaideployment, input=text)["data"][0]["embedding"]
+
===========changed ref 2===========
# module: scripts.prepdocs
+ def create_sections(filename, page_map, use_vectors):
- def create_sections(filename, page_map):
file_id = filename_to_id(filename)
+ for i, (content, pagenum) in enumerate(split_text(page_map)):
- for i, (section, pagenum) in enumerate(split_text(page_map)):
+ section = {
- yield {
"id": f"{file_id}-page-{i}",
+ "content": content,
- "content": section,
"category": args.category,
"sourcepage": blob_name_from_file_page(filename, pagenum),
"sourcefile": filename
}
+ if use_vectors:
+ section["embedding"] = compute_embedding(content)
+ yield section
===========changed ref 3===========
# module: scripts.prepdocs
def create_search_index():
if args.verbose: print(f"Ensuring search index {args.index} exists")
index_client = SearchIndexClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
credential=search_creds)
if args.index not in index_client.list_index_names():
index = SearchIndex(
name=args.index,
fields=[
SimpleField(name="id", type="Edm.String", key=True),
SearchableField(name="content", type="Edm.String", analyzer_name="en.microsoft"),
+ SearchField(name="embedding", type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
+ hidden=False, searchable=True, filterable=False, sortable=False, facetable=False,
+ vector_search_dimensions=1536, vector_search_configuration="default"),
SimpleField(name="category", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcepage", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcefile", type="Edm.String", filterable=True, facetable=True)
],
semantic_settings=SemanticSettings(
configurations=[SemanticConfiguration(
name='default',
prioritized_fields=PrioritizedFields(
+ title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))]),
- title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))])
+ vector_search=VectorSearch(
+ algorithm_configurations=[
+ VectorSearchAlgorithmConfiguration(
+ name="default",
+ kind="hnsw",
+ hnsw_parameters=HnswParameters(metric="cosine")
+ )
+ ]
+ )
+ )
- )
if args.verbose: print(f"Creating {args.index} search index")
index</s>
===========changed ref 4===========
# module: scripts.prepdocs
def create_search_index():
# offset: 1
<s> <add> )
- )
if args.verbose: print(f"Creating {args.index} search index")
index_client.create_index(index)
else:
if args.verbose: print(f"Search index {args.index} already exists")
===========changed ref 5===========
# module: scripts.prepdocs
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description="Prepare documents by extracting content from PDFs, splitting content into sections, uploading to blob storage, and indexing in a search index.",
epilog="Example: prepdocs.py '..\data\*' --storageaccount myaccount --container mycontainer --searchservice mysearch --index myindex -v"
)
parser.add_argument("files", help="Files to be processed")
parser.add_argument("--category", help="Value for the category field in the search index for all sections indexed in this run")
parser.add_argument("--skipblobs", action="store_true", help="Skip uploading individual pages to Azure Blob Storage")
parser.add_argument("--storageaccount", help="Azure Blob Storage account name")
parser.add_argument("--container", help="Azure Blob Storage container name")
parser.add_argument("--storagekey", required=False, help="Optional. Use this Azure Blob Storage account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--tenantid", required=False, help="Optional. Use this to define the Azure directory where to authenticate)")
parser.add_argument("--searchservice", help="Name of the Azure Cognitive Search service where content should be indexed (must exist already)")
parser.add_argument("--index", help="Name of the Azure Cognitive Search index where content should be indexed (will be created if it doesn't exist)")
parser.add_argument("--searchkey", required=False, help="Optional. Use this Azure Cognitive Search account key instead of the current user identity to login (use az login to set current user for Azure)")
+ parser.add_argument("--openaiservice", help="Name of the Azure OpenAI service used to compute embeddings")
+ parser.add_argument("--openaideployment", help="Name of the Azure OpenAI model deployment for an embedding model ('text-embedding-ada-002' recommended)")
+ parser.add_argument("--novectors", action="store_true", help="Don't</s>
===========changed ref 6===========
# module: scripts.prepdocs
# offset: 1
<s>-002' recommended)")
+ parser.add_argument("--novectors", action="store_true", help="Don't compute embeddings for the sections (e.g. don't call the OpenAI embeddings API during indexing)")
+ parser.add_argument("--openaikey", required=False, help="Optional. Use this Azure OpenAI account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--remove", action="store_true", help="Remove references to this document from blob storage and the search index")
parser.add_argument("--removeall", action="store_true", help="Remove all blobs from blob storage and documents from the search index")
parser.add_argument("--localpdfparser", action="store_true", help="Use PyPdf local PDF parser (supports only digital PDFs) instead of Azure Form Recognizer service to extract text, tables and layout from the documents")
parser.add_argument("--formrecognizerservice", required=False, help="Optional. Name of the Azure Form Recognizer service which will be used to extract text, tables and layout from the documents (must exist already)")
parser.add_argument("--formrecognizerkey", required=False, help="Optional. Use this Azure Form Recognizer account key instead of the current user identity to login (use az login to set current user for Azure)")
parser.add_argument("--verbose", "-v", action="store_true", help="Verbose output")
args = parser.parse_args()
# Use the current user identity to connect to Azure services unless a key is explicitly set for any of them
azd_credential = AzureDeveloperCliCredential() if args.tenantid == None else AzureDeveloperCliCredential(tenant_id=args.tenantid, process_timeout=60)
default_creds = azd_credential if args.searchkey == None or args.storagekey == None else None
search_creds = default_creds if args.searchkey</s>
|
app.backend.approaches.readretrieveread/ReadRetrieveReadApproach.retrieve
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<0>:<add> has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
<add> has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
<add> use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
<del> use_semantic_captions = True if overrides.get("semantic_captions") else False
<5>:<add> # If retrieval mode includes vectors, compute an embedding for the query
<add> if has_vector:
<add> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
<add> else:
<add> query_vector = None
<add>
<add> # Only keep the text query if the retrieval mode uses text, otherwise drop it
<add> if not has_text:
<add> query_text = None
<add>
<add> # Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
<add> if overrides.get("semantic_ranker") and has_text:
<del> if overrides.get("semantic_ranker"):
<6>:<add> r = self.search_client
|
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, q: str, overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> if overrides.get("semantic_ranker"):
<6> r = self.search_client.search(q,
<7> filter=filter,
<8> query_type=QueryType.SEMANTIC,
<9> query_language="en-us",
<10> query_speller="lexicon",
<11> semantic_configuration_name="default",
<12> top = top,
<13> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<14> else:
<15> r = self.search_client.search(q, filter=filter, top=top)
<16> if use_semantic_captions:
<17> self.results = [doc[self.sourcepage_field] + ":" + nonewlines(" -.- ".join([c.text for c in doc['@search.captions']])) for doc in r]
<18> else:
<19> self.results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:250]) for doc in r]
<20> content = "\n".join(self.results)
<21> return content
<22>
|
===========unchanged ref 0===========
at: app.backend.approaches.readretrieveread.ReadRetrieveReadApproach
template_prefix = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " \
"Answer the question using only the data provided in the information sources below. " \
"For tabular information return it as an html table. Do not return markdown format. " \
"Each source has a name followed by colon and the actual data, quote the source name for each piece of data you use in the response. " \
"For example, if the question is \"What color is the sky?\" and one of the information sources says \"info123: the sky is blue whenever it's not cloudy\", then answer with \"The sky is blue [info123]\" " \
"It's important to strictly follow the format where the name of the source is in square brackets at the end of the sentence, and only up to the prefix before the colon (\":\"). " \
"If there are multiple sources, cite each one in their own square brackets. For example, use \"[info343][ref-76]\" and not \"[info343,ref-76]\". " \
"Never quote tool names as sources." \
"If you cannot answer using the sources below, say that you don't know. " \
"\n\nYou can access to the following tools:"
template_suffix = """
Begin!
Question: {input}
Thought: {agent_scratchpad}"""
CognitiveSearchToolDescription = "useful for searching the Microsoft employee benefits information such as healthcare plans, retirement plans, etc."
at: app.backend.approaches.readretrieveread.ReadRetrieveReadApproach.__init__
self.search_client = search_client
self.embedding_deployment = embedding_deployment
===========unchanged ref 1===========
at: openai.api_resources.embedding
Embedding(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.embedding.Embedding
OBJECT_NAME = "embeddings"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 1===========
# module: scripts.prepdocs
+ def before_retry_sleep(retry_state):
+ if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
+
===========changed ref 2===========
# module: scripts.prepdocs
+ @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(15), before_sleep=before_retry_sleep)
+ def compute_embedding(text):
+ return openai.Embedding.create(engine=args.openaideployment, input=text)["data"][0]["embedding"]
+
===========changed ref 3===========
# module: scripts.prepdocs
+ def create_sections(filename, page_map, use_vectors):
- def create_sections(filename, page_map):
file_id = filename_to_id(filename)
+ for i, (content, pagenum) in enumerate(split_text(page_map)):
- for i, (section, pagenum) in enumerate(split_text(page_map)):
+ section = {
- yield {
"id": f"{file_id}-page-{i}",
+ "content": content,
- "content": section,
"category": args.category,
"sourcepage": blob_name_from_file_page(filename, pagenum),
"sourcefile": filename
}
+ if use_vectors:
+ section["embedding"] = compute_embedding(content)
+ yield section
===========changed ref 4===========
# module: scripts.prepdocs
def create_search_index():
if args.verbose: print(f"Ensuring search index {args.index} exists")
index_client = SearchIndexClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
credential=search_creds)
if args.index not in index_client.list_index_names():
index = SearchIndex(
name=args.index,
fields=[
SimpleField(name="id", type="Edm.String", key=True),
SearchableField(name="content", type="Edm.String", analyzer_name="en.microsoft"),
+ SearchField(name="embedding", type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
+ hidden=False, searchable=True, filterable=False, sortable=False, facetable=False,
+ vector_search_dimensions=1536, vector_search_configuration="default"),
SimpleField(name="category", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcepage", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcefile", type="Edm.String", filterable=True, facetable=True)
],
semantic_settings=SemanticSettings(
configurations=[SemanticConfiguration(
name='default',
prioritized_fields=PrioritizedFields(
+ title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))]),
- title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))])
+ vector_search=VectorSearch(
+ algorithm_configurations=[
+ VectorSearchAlgorithmConfiguration(
+ name="default",
+ kind="hnsw",
+ hnsw_parameters=HnswParameters(metric="cosine")
+ )
+ ]
+ )
+ )
- )
if args.verbose: print(f"Creating {args.index} search index")
index</s>
===========changed ref 5===========
# module: scripts.prepdocs
def create_search_index():
# offset: 1
<s> <add> )
- )
if args.verbose: print(f"Creating {args.index} search index")
index_client.create_index(index)
else:
if args.verbose: print(f"Search index {args.index} already exists")
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<3>:<add> self.embedding_deployment = embedding_deployment
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
<0> self.search_client = search_client
<1> self.chatgpt_deployment = chatgpt_deployment
<2> self.chatgpt_model = chatgpt_model
<3> self.sourcepage_field = sourcepage_field
<4> self.content_field = content_field
<5> self.chatgpt_token_limit = get_token_limit(chatgpt_model)
<6>
|
===========unchanged ref 0===========
at: core.modelhelper
get_token_limit(model_id: str) -> int
===========changed ref 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# Chat roles
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
+ For tabular information return it as an html table. Do not return markdown format. If the question is not in English, answer in the language used in the question.
- For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
+ Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
- Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
+ Try not to repeat questions that have already been asked.
- Try not to repeat questions that have already been asked.
+ Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'</s>
===========changed ref 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
# offset: 1
<s> have already been asked.
+ Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
- Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
+ Generate a search query based on the conversation and the new question.
- Generate a search query based on the conversation and the new question.
+ Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
- Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
+ Do not include any text inside [] or <<>> in the search query terms.
- Do not include any text inside [] or <<>> in the search query terms.
+ Do not include any special characters like '+'.
- Do not include any special characters like '+'.
+ If the question is not in English, translate the question to English before generating the search query.
- If the question is not in English, translate the question to English before generating the search query.
-
- Search Query:
+ If you cannot generate a search query, return just the number 0.
"""
query_prompt_few_shots = [
{'role' : USER, 'content' : 'What are my health plans?' },
{'role' : ASSISTANT, 'content' : 'Show available health plans' },
{'role' : USER, 'content' : 'does my plan cover cardio?' },
{'role' : ASSISTANT, 'content' : 'Health plan cardio coverage' }
]
===========changed ref 2===========
# module: scripts.prepdocs
+ def before_retry_sleep(retry_state):
+ if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
+
===========changed ref 3===========
# module: scripts.prepdocs
+ @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(15), before_sleep=before_retry_sleep)
+ def compute_embedding(text):
+ return openai.Embedding.create(engine=args.openaideployment, input=text)["data"][0]["embedding"]
+
===========changed ref 4===========
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 5===========
# module: scripts.prepdocs
+ def create_sections(filename, page_map, use_vectors):
- def create_sections(filename, page_map):
file_id = filename_to_id(filename)
+ for i, (content, pagenum) in enumerate(split_text(page_map)):
- for i, (section, pagenum) in enumerate(split_text(page_map)):
+ section = {
- yield {
"id": f"{file_id}-page-{i}",
+ "content": content,
- "content": section,
"category": args.category,
"sourcepage": blob_name_from_file_page(filename, pagenum),
"sourcefile": filename
}
+ if use_vectors:
+ section["embedding"] = compute_embedding(content)
+ yield section
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<0>:<add> has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
<add> has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
<add> use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
<del> use_semantic_captions = True if overrides.get("semantic_captions") else False
<25>:<add> query_text = chat_completion.choices[0].message.content
<del> q = chat_completion.choices[0].message.content
<26>:<add> if query_text.strip() == "0":
<add> query_text = history[-1]["user"] # Use the last user input if we failed to generate a better query
<28>:<add>
<add> # If retrieval mode includes vectors, compute an embedding for the query
<add> if has_vector:
<add> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
<add> else:
<add> query_vector = None
<add>
<add> # Only keep
|
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> user_q = 'Generate search query for: ' + history[-1]["user"]
<6>
<7> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<8> messages = self.get_messages_from_history(
<9> self.query_prompt_template,
<10> self.chatgpt_model,
<11> history,
<12> user_q,
<13> self.query_prompt_few_shots,
<14> self.chatgpt_token_limit - len(user_q)
<15> )
<16>
<17> chat_completion = openai.ChatCompletion.create(
<18> deployment_id=self.chatgpt_deployment,
<19> model=self.chatgpt_model,
<20> messages=messages,
<21> temperature=0.0,
<22> max_tokens=32,
<23> n=1)
<24>
<25> q = chat_completion.choices[0].message.content
<26>
<27> # STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
<28> if overrides.get("semantic_ranker"):
<29> r = self.search_client.search(q,
<30> filter=filter,
<31> query_type=QueryType.SEMANTIC,
<32> query_language="en-us",
<33> query_speller="lexicon",
<34> semantic_configuration_name="default",
<35> top=top,
<36> query_caption="extractive</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
else:
r = self.search_client.search(q, filter=filter, top=top)
if use_semantic_captions:
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_override")
if prompt_override is None:
system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
else:
system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
# latest conversation
user_content = history[-1]["user"] + " \nSources:" + content
messages = self.get_messages_from_history(
system_message,
self.chatgpt_model,
history,
user_content,
max_tokens=self.chatg</s>
===========below chunk 1===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 2
<s>
self.chatgpt_model,
history,
user_content,
max_tokens=self.chatgpt_token_limit)
chat_completion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chat_content = chat_completion.choices[0].message.content
msg_to_display = '\n\n'.join([str(message) for message in messages])
return {"data_points": results, "answer": chat_content, "thoughts": f"Searched for:<br>{q}<br><br>Conversations:<br>" + msg_to_display.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
Do not include any special characters like '+'.
If the question is not in English, translate the question to English before generating the search query.
Search Query:
"""
query_prompt_few_shots = [
{'role' : USER, 'content' : 'What are my health plans?' },
{'role' : ASSISTANT, 'content' : 'Show available health plans' },
{'role' : USER, 'content' : 'does my plan cover cardio?' },
{'role' : ASSISTANT, 'content' : 'Health plan cardio coverage' }
]
get_messages_from_history(system_prompt: str, model_id: str, history: Sequence[dict[str, str]], user_conv: str, few_shots=[], max_tokens: int=4096) -> []
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach.__init__
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
self.sourcepage_field = sourcepage_field
self.content_field = content_field
self.chatgpt_token_limit = get_token_limit(chatgpt_model)
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 2===========
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing
Sequence = _alias(collections.abc.Sequence, 1)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
app.backend.approaches.retrievethenread/RetrieveThenReadApproach.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<3>:<add> self.embedding_deployment = embedding_deployment
|
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
<0> self.search_client = search_client
<1> self.openai_deployment = openai_deployment
<2> self.chatgpt_model = chatgpt_model
<3> self.sourcepage_field = sourcepage_field
<4> self.content_field = content_field
<5>
|
===========changed ref 0===========
# module: scripts.prepdocs
+ def before_retry_sleep(retry_state):
+ if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
+
===========changed ref 1===========
# module: scripts.prepdocs
+ @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(15), before_sleep=before_retry_sleep)
+ def compute_embedding(text):
+ return openai.Embedding.create(engine=args.openaideployment, input=text)["data"][0]["embedding"]
+
===========changed ref 2===========
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 3===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
self.chatgpt_token_limit = get_token_limit(chatgpt_model)
===========changed ref 4===========
# module: scripts.prepdocs
+ def create_sections(filename, page_map, use_vectors):
- def create_sections(filename, page_map):
file_id = filename_to_id(filename)
+ for i, (content, pagenum) in enumerate(split_text(page_map)):
- for i, (section, pagenum) in enumerate(split_text(page_map)):
+ section = {
- yield {
"id": f"{file_id}-page-{i}",
+ "content": content,
- "content": section,
"category": args.category,
"sourcepage": blob_name_from_file_page(filename, pagenum),
"sourcefile": filename
}
+ if use_vectors:
+ section["embedding"] = compute_embedding(content)
+ yield section
===========changed ref 5===========
# module: scripts.prepdocs
def create_search_index():
if args.verbose: print(f"Ensuring search index {args.index} exists")
index_client = SearchIndexClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
credential=search_creds)
if args.index not in index_client.list_index_names():
index = SearchIndex(
name=args.index,
fields=[
SimpleField(name="id", type="Edm.String", key=True),
SearchableField(name="content", type="Edm.String", analyzer_name="en.microsoft"),
+ SearchField(name="embedding", type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
+ hidden=False, searchable=True, filterable=False, sortable=False, facetable=False,
+ vector_search_dimensions=1536, vector_search_configuration="default"),
SimpleField(name="category", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcepage", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcefile", type="Edm.String", filterable=True, facetable=True)
],
semantic_settings=SemanticSettings(
configurations=[SemanticConfiguration(
name='default',
prioritized_fields=PrioritizedFields(
+ title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))]),
- title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))])
+ vector_search=VectorSearch(
+ algorithm_configurations=[
+ VectorSearchAlgorithmConfiguration(
+ name="default",
+ kind="hnsw",
+ hnsw_parameters=HnswParameters(metric="cosine")
+ )
+ ]
+ )
+ )
- )
if args.verbose: print(f"Creating {args.index} search index")
index</s>
===========changed ref 6===========
# module: scripts.prepdocs
def create_search_index():
# offset: 1
<s> <add> )
- )
if args.verbose: print(f"Creating {args.index} search index")
index_client.create_index(index)
else:
if args.verbose: print(f"Search index {args.index} already exists")
===========changed ref 7===========
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, q: str, overrides: dict[str, Any]) -> Any:
+ has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
+ has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
+ use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
- use_semantic_captions = True if overrides.get("semantic_captions") else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
+ # If retrieval mode includes vectors, compute an embedding for the query
+ if has_vector:
+ query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
+ else:
+ query_vector = None
+
+ # Only keep the text query if the retrieval mode uses text, otherwise drop it
+ if not has_text:
+ query_text = None
+
+ # Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
+ if overrides.get("semantic_ranker") and has_text:
- if overrides.get("semantic_ranker"):
+ r = self.search_client.search(query_text,
- r = self.search_client.search(q,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
top = top,
+ </s>
|
app.backend.approaches.retrievethenread/RetrieveThenReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<0>:<add> has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
<add> has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
<add> use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
<del> use_semantic_captions = True if overrides.get("semantic_captions") else False
<5>:<add> # If retrieval mode includes vectors, compute an embedding for the query
<add> if has_vector:
<add> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=q)["data"][0]["embedding"]
<add> else:
<add> query_vector = None
<add>
<add> # Only keep the text query if the retrieval mode uses text, otherwise drop it
<add> query_text = q if has_text else None
<add>
<add> # Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
<add> if overrides.get("semantic_ranker") and has_text:
<del> if overrides.get("semantic_ranker"):
<6>:<add> r = self.search_client.search(query_
|
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
def run(self, q: str, overrides: dict[str, Any]) -> Any:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> if overrides.get("semantic_ranker"):
<6> r = self.search_client.search(q,
<7> filter=filter,
<8> query_type=QueryType.SEMANTIC,
<9> query_language="en-us",
<10> query_speller="lexicon",
<11> semantic_configuration_name="default",
<12> top=top,
<13> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<14> else:
<15> r = self.search_client.search(q, filter=filter, top=top)
<16> if use_semantic_captions:
<17> results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
<18> else:
<19> results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
<20> content = "\n".join(results)
<21>
<22> message_builder = MessageBuilder(overrides.get("prompt_template") or self.system_chat_template, self.chatgpt_model);
<23>
<24> # add user question
<25> user_content = q + "\n" + "Sources:\n {content}".format(content=content)
<26> message_builder.append_message('user', user_content)
<27>
<28> # Add shots/samples. This helps model to</s>
|
===========below chunk 0===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
message_builder.append_message('assistant', self.answer)
message_builder.append_message('user', self.question)
messages = message_builder.messages
chat_completion = openai.ChatCompletion.create(
deployment_id=self.openai_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.3,
max_tokens=1024,
n=1)
return {"data_points": results, "answer": chat_completion.choices[0].message.content, "thoughts": f"Question:<br>{q}<br><br>Prompt:<br>" + '\n\n'.join([str(message) for message in messages])}
===========unchanged ref 0===========
at: app.backend.approaches.retrievethenread.RetrieveThenReadApproach
system_chat_template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
"If you cannot answer using the sources below, say you don't know. Use below example to answer"
question = """
'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
"""
answer = "In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf]."
at: app.backend.approaches.retrievethenread.RetrieveThenReadApproach.__init__
self.search_client = search_client
self.openai_deployment = openai_deployment
self.chatgpt_model = chatgpt_model
===========unchanged ref 1===========
self.sourcepage_field = sourcepage_field
self.content_field = content_field
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: core.messagebuilder
MessageBuilder(system_content: str, chatgpt_model: str)
at: core.messagebuilder.MessageBuilder
append_message(role: str, content: str, index: int=1)
at: core.messagebuilder.MessageBuilder.__init__
self.messages = [{'role': 'system', 'content': system_content}]
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
self.chatgpt_model = chatgpt_model
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 1===========
# module: scripts.prepdocs
+ def before_retry_sleep(retry_state):
+ if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
+
===========changed ref 2===========
# module: scripts.prepdocs
+ @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(15), before_sleep=before_retry_sleep)
+ def compute_embedding(text):
+ return openai.Embedding.create(engine=args.openaideployment, input=text)["data"][0]["embedding"]
+
===========changed ref 3===========
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 4===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
self.chatgpt_token_limit = get_token_limit(chatgpt_model)
|
app.backend.approaches.readdecomposeask/ReadDecomposeAsk.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<2>:<add> self.embedding_deployment = embedding_deployment
|
# module: app.backend.approaches.readdecomposeask
class ReadDecomposeAsk(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
<0> self.search_client = search_client
<1> self.openai_deployment = openai_deployment
<2> self.sourcepage_field = sourcepage_field
<3> self.content_field = content_field
<4>
|
===========changed ref 0===========
# module: scripts.prepdocs
+ def before_retry_sleep(retry_state):
+ if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
+
===========changed ref 1===========
# module: scripts.prepdocs
+ @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(15), before_sleep=before_retry_sleep)
+ def compute_embedding(text):
+ return openai.Embedding.create(engine=args.openaideployment, input=text)["data"][0]["embedding"]
+
===========changed ref 2===========
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 3===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
self.chatgpt_model = chatgpt_model
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 4===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
self.chatgpt_token_limit = get_token_limit(chatgpt_model)
===========changed ref 5===========
# module: scripts.prepdocs
+ def create_sections(filename, page_map, use_vectors):
- def create_sections(filename, page_map):
file_id = filename_to_id(filename)
+ for i, (content, pagenum) in enumerate(split_text(page_map)):
- for i, (section, pagenum) in enumerate(split_text(page_map)):
+ section = {
- yield {
"id": f"{file_id}-page-{i}",
+ "content": content,
- "content": section,
"category": args.category,
"sourcepage": blob_name_from_file_page(filename, pagenum),
"sourcefile": filename
}
+ if use_vectors:
+ section["embedding"] = compute_embedding(content)
+ yield section
===========changed ref 6===========
# module: scripts.prepdocs
def create_search_index():
if args.verbose: print(f"Ensuring search index {args.index} exists")
index_client = SearchIndexClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
credential=search_creds)
if args.index not in index_client.list_index_names():
index = SearchIndex(
name=args.index,
fields=[
SimpleField(name="id", type="Edm.String", key=True),
SearchableField(name="content", type="Edm.String", analyzer_name="en.microsoft"),
+ SearchField(name="embedding", type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
+ hidden=False, searchable=True, filterable=False, sortable=False, facetable=False,
+ vector_search_dimensions=1536, vector_search_configuration="default"),
SimpleField(name="category", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcepage", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcefile", type="Edm.String", filterable=True, facetable=True)
],
semantic_settings=SemanticSettings(
configurations=[SemanticConfiguration(
name='default',
prioritized_fields=PrioritizedFields(
+ title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))]),
- title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))])
+ vector_search=VectorSearch(
+ algorithm_configurations=[
+ VectorSearchAlgorithmConfiguration(
+ name="default",
+ kind="hnsw",
+ hnsw_parameters=HnswParameters(metric="cosine")
+ )
+ ]
+ )
+ )
- )
if args.verbose: print(f"Creating {args.index} search index")
index</s>
===========changed ref 7===========
# module: scripts.prepdocs
def create_search_index():
# offset: 1
<s> <add> )
- )
if args.verbose: print(f"Creating {args.index} search index")
index_client.create_index(index)
else:
if args.verbose: print(f"Search index {args.index} already exists")
|
app.backend.approaches.readdecomposeask/ReadDecomposeAsk.search
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
85791db8b36483aaca35cb288e20706391323b14
|
Vector search (#424)
|
<0>:<add> has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
<add> has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
<add> use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
<del> use_semantic_captions = True if overrides.get("semantic_captions") else False
<5>:<add> # If retrieval mode includes vectors, compute an embedding for the query
<add> if has_vector:
<add> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
<add> else:
<add> query_vector = None
<add>
<add> # Only keep the text query if the retrieval mode uses text, otherwise drop it
<add> if not has_text:
<add> query_text = None
<add>
<add> if overrides.get("semantic_ranker") and has_text:
<del> if overrides.get("semantic_ranker"):
<6>:<add> r = self.search_client.search(query_text,
<del> r = self.search_client.search(q,
<12>:<add> top=top,
|
# module: app.backend.approaches.readdecomposeask
class ReadDecomposeAsk(Approach):
+ def search(self, query_text: str, overrides: dict[str, Any]) -> str:
- def search(self, q: str, overrides: dict[str, Any]) -> str:
<0> use_semantic_captions = True if overrides.get("semantic_captions") else False
<1> top = overrides.get("top") or 3
<2> exclude_category = overrides.get("exclude_category") or None
<3> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<4>
<5> if overrides.get("semantic_ranker"):
<6> r = self.search_client.search(q,
<7> filter=filter,
<8> query_type=QueryType.SEMANTIC,
<9> query_language="en-us",
<10> query_speller="lexicon",
<11> semantic_configuration_name="default",
<12> top = top,
<13> query_caption="extractive|highlight-false" if use_semantic_captions else None)
<14> else:
<15> r = self.search_client.search(q, filter=filter, top=top)
<16> if use_semantic_captions:
<17> self.results = [doc[self.sourcepage_field] + ":" + nonewlines(" . ".join([c.text for c in doc['@search.captions'] ])) for doc in r]
<18> else:
<19> self.results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:500]) for doc in r]
<20> return "\n".join(self.results)
<21>
|
===========unchanged ref 0===========
at: app.backend.approaches.readdecomposeask.ReadDecomposeAsk.__init__
self.search_client = search_client
self.embedding_deployment = embedding_deployment
at: openai.api_resources.embedding
Embedding(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.embedding.Embedding
OBJECT_NAME = "embeddings"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.readdecomposeask
class ReadDecomposeAsk(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 1===========
# module: scripts.prepdocs
+ def before_retry_sleep(retry_state):
+ if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
+
===========changed ref 2===========
# module: scripts.prepdocs
+ @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(15), before_sleep=before_retry_sleep)
+ def compute_embedding(text):
+ return openai.Embedding.create(engine=args.openaideployment, input=text)["data"][0]["embedding"]
+
===========changed ref 3===========
# module: app.backend.approaches.readretrieveread
class ReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 4===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, openai_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.openai_deployment = openai_deployment
self.chatgpt_model = chatgpt_model
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
===========changed ref 5===========
# module: app.backend.approaches.chatreadretrieveread
class ChatReadRetrieveReadApproach(Approach):
+ def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, embedding_deployment: str, sourcepage_field: str, content_field: str):
- def __init__(self, search_client: SearchClient, chatgpt_deployment: str, chatgpt_model: str, sourcepage_field: str, content_field: str):
self.search_client = search_client
self.chatgpt_deployment = chatgpt_deployment
self.chatgpt_model = chatgpt_model
+ self.embedding_deployment = embedding_deployment
self.sourcepage_field = sourcepage_field
self.content_field = content_field
self.chatgpt_token_limit = get_token_limit(chatgpt_model)
===========changed ref 6===========
# module: scripts.prepdocs
+ def create_sections(filename, page_map, use_vectors):
- def create_sections(filename, page_map):
file_id = filename_to_id(filename)
+ for i, (content, pagenum) in enumerate(split_text(page_map)):
- for i, (section, pagenum) in enumerate(split_text(page_map)):
+ section = {
- yield {
"id": f"{file_id}-page-{i}",
+ "content": content,
- "content": section,
"category": args.category,
"sourcepage": blob_name_from_file_page(filename, pagenum),
"sourcefile": filename
}
+ if use_vectors:
+ section["embedding"] = compute_embedding(content)
+ yield section
===========changed ref 7===========
# module: scripts.prepdocs
def create_search_index():
if args.verbose: print(f"Ensuring search index {args.index} exists")
index_client = SearchIndexClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
credential=search_creds)
if args.index not in index_client.list_index_names():
index = SearchIndex(
name=args.index,
fields=[
SimpleField(name="id", type="Edm.String", key=True),
SearchableField(name="content", type="Edm.String", analyzer_name="en.microsoft"),
+ SearchField(name="embedding", type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
+ hidden=False, searchable=True, filterable=False, sortable=False, facetable=False,
+ vector_search_dimensions=1536, vector_search_configuration="default"),
SimpleField(name="category", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcepage", type="Edm.String", filterable=True, facetable=True),
SimpleField(name="sourcefile", type="Edm.String", filterable=True, facetable=True)
],
semantic_settings=SemanticSettings(
configurations=[SemanticConfiguration(
name='default',
prioritized_fields=PrioritizedFields(
+ title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))]),
- title_field=None, prioritized_content_fields=[SemanticField(field_name='content')]))])
+ vector_search=VectorSearch(
+ algorithm_configurations=[
+ VectorSearchAlgorithmConfiguration(
+ name="default",
+ kind="hnsw",
+ hnsw_parameters=HnswParameters(metric="cosine")
+ )
+ ]
+ )
+ )
- )
if args.verbose: print(f"Creating {args.index} search index")
index</s>
===========changed ref 8===========
# module: scripts.prepdocs
def create_search_index():
# offset: 1
<s> <add> )
- )
if args.verbose: print(f"Creating {args.index} search index")
index_client.create_index(index)
else:
if args.verbose: print(f"Search index {args.index} already exists")
|
app.backend.core.modelhelper/get_token_limit
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9b4b65bab522ad0c9a1e445f6b10e3cc6daac491
|
Add more tests for core/ folder (#430)
|
<1>:<add> raise ValueError("Expected model gpt-35-turbo and above")
<del> raise ValueError("Expected Model Gpt-35-turbo and above")
|
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
<0> if model_id not in MODELS_2_TOKEN_LIMITS:
<1> raise ValueError("Expected Model Gpt-35-turbo and above")
<2> return MODELS_2_TOKEN_LIMITS.get(model_id)
<3>
|
===========unchanged ref 0===========
at: app.backend.core.modelhelper
MODELS_2_TOKEN_LIMITS = {
"gpt-35-turbo": 4000,
"gpt-3.5-turbo": 4000,
"gpt-35-turbo-16k": 16000,
"gpt-3.5-turbo-16k": 16000,
"gpt-4": 8100,
"gpt-4-32k": 32000
}
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
app.backend.core.modelhelper/get_oai_chatmodel_tiktok
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9b4b65bab522ad0c9a1e445f6b10e3cc6daac491
|
Add more tests for core/ folder (#430)
|
<0>:<add> message = "Expected Azure OpenAI ChatGPT model name"
<1>:<add> raise ValueError(message)
<add> if aoaimodel not in AOAI_2_OAI and aoaimodel not in MODELS_2_TOKEN_LIMITS:
<add> raise ValueError(message)
<add> return AOAI_2_OAI.get(aoaimodel) or aoaimodel
<del> raise ValueError("Expected AOAI chatGPT model name")
<3>:<del> return AOAI_2_OAI.get(aoaimodel)
<4>:<del>
|
# module: app.backend.core.modelhelper
def get_oai_chatmodel_tiktok(aoaimodel: str) -> str:
<0> if aoaimodel == "" or aoaimodel is None:
<1> raise ValueError("Expected AOAI chatGPT model name")
<2>
<3> return AOAI_2_OAI.get(aoaimodel)
<4>
|
===========unchanged ref 0===========
at: app.backend.core.modelhelper
MODELS_2_TOKEN_LIMITS = {
"gpt-35-turbo": 4000,
"gpt-3.5-turbo": 4000,
"gpt-35-turbo-16k": 16000,
"gpt-3.5-turbo-16k": 16000,
"gpt-4": 8100,
"gpt-4-32k": 32000
}
AOAI_2_OAI = {
"gpt-35-turbo": "gpt-3.5-turbo",
"gpt-35-turbo-16k": "gpt-3.5-turbo-16k"
}
===========changed ref 0===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
+ raise ValueError("Expected model gpt-35-turbo and above")
- raise ValueError("Expected Model Gpt-35-turbo and above")
return MODELS_2_TOKEN_LIMITS.get(model_id)
|
app.backend.app/ask
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
52abf79fdf545fc29c5c9159b88a6ff9010f4ed2
|
Ensure openai token (#262)
|
<0>:<del> ensure_openai_token()
|
# module: app.backend.app
@app.route("/ask", methods=["POST"])
def ask():
<0> ensure_openai_token()
<1> if not request.json:
<2> return jsonify({"error": "request must be json"}), 400
<3> approach = request.json["approach"]
<4> try:
<5> impl = ask_approaches.get(approach)
<6> if not impl:
<7> return jsonify({"error": "unknown approach"}), 400
<8> r = impl.run(request.json["question"], request.json.get("overrides") or {})
<9> return jsonify(r)
<10> except Exception as e:
<11> logging.exception("Exception in /ask")
<12> return jsonify({"error": str(e)}), 500
<13>
|
===========unchanged ref 0===========
at: app.backend.app
ask_approaches = {
"rtr": RetrieveThenReadApproach(search_client, AZURE_OPENAI_CHATGPT_DEPLOYMENT, AZURE_OPENAI_CHATGPT_MODEL, AZURE_OPENAI_EMB_DEPLOYMENT, KB_FIELDS_SOURCEPAGE, KB_FIELDS_CONTENT),
"rrr": ReadRetrieveReadApproach(search_client, AZURE_OPENAI_GPT_DEPLOYMENT, AZURE_OPENAI_EMB_DEPLOYMENT, KB_FIELDS_SOURCEPAGE, KB_FIELDS_CONTENT),
"rda": ReadDecomposeAsk(search_client, AZURE_OPENAI_GPT_DEPLOYMENT, AZURE_OPENAI_EMB_DEPLOYMENT, KB_FIELDS_SOURCEPAGE, KB_FIELDS_CONTENT)
}
app = Flask(__name__)
at: approaches.readdecomposeask.ReadDecomposeAsk
run(q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 1===========
at: approaches.readretrieveread.ReadRetrieveReadApproach
template_prefix = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " \
"Answer the question using only the data provided in the information sources below. " \
"For tabular information return it as an html table. Do not return markdown format. " \
"Each source has a name followed by colon and the actual data, quote the source name for each piece of data you use in the response. " \
"For example, if the question is \"What color is the sky?\" and one of the information sources says \"info123: the sky is blue whenever it's not cloudy\", then answer with \"The sky is blue [info123]\" " \
"It's important to strictly follow the format where the name of the source is in square brackets at the end of the sentence, and only up to the prefix before the colon (\":\"). " \
"If there are multiple sources, cite each one in their own square brackets. For example, use \"[info343][ref-76]\" and not \"[info343,ref-76]\". " \
"Never quote tool names as sources." \
"If you cannot answer using the sources below, say that you don't know. " \
"\n\nYou can access to the following tools:"
template_suffix = """
Begin!
Question: {input}
Thought: {agent_scratchpad}"""
CognitiveSearchToolDescription = "useful for searching the Microsoft employee benefits information such as healthcare plans, retirement plans, etc."
run(q: str, overrides: dict[str, Any]) -> Any
===========unchanged ref 2===========
at: approaches.retrievethenread.RetrieveThenReadApproach
system_chat_template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
"If you cannot answer using the sources below, say you don't know. Use below example to answer"
question = """
'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
"""
answer = "In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf]."
run(q: str, overrides: dict[str, Any]) -> Any
at: logging
exception(msg: Any, *args: Any, exc_info: _ExcInfoType=..., stack_info: bool=..., extra: Optional[Dict[str, Any]]=..., **kwargs: Any) -> None
===========unchanged ref 3===========
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
|
app.backend.app/chat
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
52abf79fdf545fc29c5c9159b88a6ff9010f4ed2
|
Ensure openai token (#262)
|
<0>:<del> ensure_openai_token()
|
# module: app.backend.app
@app.route("/chat", methods=["POST"])
def chat():
<0> ensure_openai_token()
<1> if not request.json:
<2> return jsonify({"error": "request must be json"}), 400
<3> approach = request.json["approach"]
<4> try:
<5> impl = chat_approaches.get(approach)
<6> if not impl:
<7> return jsonify({"error": "unknown approach"}), 400
<8> r = impl.run(request.json["history"], request.json.get("overrides") or {})
<9> return jsonify(r)
<10> except Exception as e:
<11> logging.exception("Exception in /chat")
<12> return jsonify({"error": str(e)}), 500
<13>
|
===========unchanged ref 0===========
at: app.backend.app
chat_approaches = {
"rrr": ChatReadRetrieveReadApproach(search_client,
AZURE_OPENAI_CHATGPT_DEPLOYMENT,
AZURE_OPENAI_CHATGPT_MODEL,
AZURE_OPENAI_EMB_DEPLOYMENT,
KB_FIELDS_SOURCEPAGE,
KB_FIELDS_CONTENT)
}
app = Flask(__name__)
at: approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format. If the question is not in English, answer in the language used in the question.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
===========unchanged ref 1===========
query_prompt_template = """Below is a history of the conversation so far, and a new question asked by the user that needs to be answered by searching in a knowledge base about employee healthcare plans and the employee handbook.
Generate a search query based on the conversation and the new question.
Do not include cited source filenames and document names e.g info.txt or doc.pdf in the search query terms.
Do not include any text inside [] or <<>> in the search query terms.
Do not include any special characters like '+'.
If the question is not in English, translate the question to English before generating the search query.
If you cannot generate a search query, return just the number 0.
"""
query_prompt_few_shots = [
{'role' : USER, 'content' : 'What are my health plans?' },
{'role' : ASSISTANT, 'content' : 'Show available health plans' },
{'role' : USER, 'content' : 'does my plan cover cardio?' },
{'role' : ASSISTANT, 'content' : 'Health plan cardio coverage' }
]
run(history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any
at: logging
exception(msg: Any, *args: Any, exc_info: _ExcInfoType=..., stack_info: bool=..., extra: Optional[Dict[str, Any]]=..., **kwargs: Any) -> None
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.app
@app.route("/ask", methods=["POST"])
def ask():
- ensure_openai_token()
if not request.json:
return jsonify({"error": "request must be json"}), 400
approach = request.json["approach"]
try:
impl = ask_approaches.get(approach)
if not impl:
return jsonify({"error": "unknown approach"}), 400
r = impl.run(request.json["question"], request.json.get("overrides") or {})
return jsonify(r)
except Exception as e:
logging.exception("Exception in /ask")
return jsonify({"error": str(e)}), 500
|
app.backend.app/ensure_openai_token
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
52abf79fdf545fc29c5c9159b88a6ff9010f4ed2
|
Ensure openai token (#262)
|
<1>:<add> if openai_token.expires_on < time.time() + 60:
<del> if openai_token.expires_on < int(time.time()) - 60:
|
# module: app.backend.app
+ @app.before_request
def ensure_openai_token():
<0> global openai_token
<1> if openai_token.expires_on < int(time.time()) - 60:
<2> openai_token = azure_credential.get_token("https://cognitiveservices.azure.com/.default")
<3> openai.api_key = openai_token.token
<4>
|
===========unchanged ref 0===========
at: app.backend.app
azure_credential = DefaultAzureCredential(exclude_shared_token_cache_credential = True)
at: openai
api_key = os.environ.get("OPENAI_API_KEY")
at: time
time() -> float
===========changed ref 0===========
# module: app.backend.app
@app.route("/chat", methods=["POST"])
def chat():
- ensure_openai_token()
if not request.json:
return jsonify({"error": "request must be json"}), 400
approach = request.json["approach"]
try:
impl = chat_approaches.get(approach)
if not impl:
return jsonify({"error": "unknown approach"}), 400
r = impl.run(request.json["history"], request.json.get("overrides") or {})
return jsonify(r)
except Exception as e:
logging.exception("Exception in /chat")
return jsonify({"error": str(e)}), 500
===========changed ref 1===========
# module: app.backend.app
@app.route("/ask", methods=["POST"])
def ask():
- ensure_openai_token()
if not request.json:
return jsonify({"error": "request must be json"}), 400
approach = request.json["approach"]
try:
impl = ask_approaches.get(approach)
if not impl:
return jsonify({"error": "unknown approach"}), 400
r = impl.run(request.json["question"], request.json.get("overrides") or {})
return jsonify(r)
except Exception as e:
logging.exception("Exception in /ask")
return jsonify({"error": str(e)}), 500
|
scripts.prepdocs/remove_blobs
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
07a97f3a74b75e6e938cec471fabd1fa7a9111f1
|
Run ruff, add to precommit (#491)
|
<4>:<add> if filename is None:
<del> if filename == None:
|
# module: scripts.prepdocs
def remove_blobs(filename):
<0> if args.verbose: print(f"Removing blobs for '{filename or '<all>'}'")
<1> blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
<2> blob_container = blob_service.get_container_client(args.container)
<3> if blob_container.exists():
<4> if filename == None:
<5> blobs = blob_container.list_blob_names()
<6> else:
<7> prefix = os.path.splitext(os.path.basename(filename))[0]
<8> blobs = filter(lambda b: re.match(f"{prefix}-\d+\.pdf", b), blob_container.list_blob_names(name_starts_with=os.path.splitext(os.path.basename(prefix))[0]))
<9> for b in blobs:
<10> if args.verbose: print(f"\tRemoving blob {b}")
<11> blob_container.delete_blob(b)
<12>
|
===========unchanged ref 0===========
at: os.path
splitext(p: AnyStr) -> Tuple[AnyStr, AnyStr]
splitext(p: _PathLike[AnyStr]) -> Tuple[AnyStr, AnyStr]
basename(p: _PathLike[AnyStr]) -> AnyStr
basename(p: AnyStr) -> AnyStr
at: re
match(pattern: AnyStr, string: AnyStr, flags: _FlagsType=...) -> Optional[Match[AnyStr]]
match(pattern: Pattern[AnyStr], string: AnyStr, flags: _FlagsType=...) -> Optional[Match[AnyStr]]
at: scripts.prepdocs
args = parser.parse_args()
storage_creds = default_creds if args.storagekey is None else args.storagekey
|
scripts.prepdocs/get_document_text
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
07a97f3a74b75e6e938cec471fabd1fa7a9111f1
|
Run ruff, add to precommit (#491)
|
# module: scripts.prepdocs
def get_document_text(filename):
<0> offset = 0
<1> page_map = []
<2> if args.localpdfparser:
<3> reader = PdfReader(filename)
<4> pages = reader.pages
<5> for page_num, p in enumerate(pages):
<6> page_text = p.extract_text()
<7> page_map.append((page_num, offset, page_text))
<8> offset += len(page_text)
<9> else:
<10> if args.verbose: print(f"Extracting text from '{filename}' using Azure Form Recognizer")
<11> form_recognizer_client = DocumentAnalysisClient(endpoint=f"https://{args.formrecognizerservice}.cognitiveservices.azure.com/", credential=formrecognizer_creds, headers={"x-ms-useragent": "azure-search-chat-demo/1.0.0"})
<12> with open(filename, "rb") as f:
<13> poller = form_recognizer_client.begin_analyze_document("prebuilt-layout", document = f)
<14> form_recognizer_results = poller.result()
<15>
<16> for page_num, page in enumerate(form_recognizer_results.pages):
<17> tables_on_page = [table for table in form_recognizer_results.tables if table.bounding_regions[0].page_number == page_num + 1]
<18>
<19> # mark all positions of the table spans in the page
<20> page_offset = page.spans[0].offset
<21> page_length = page.spans[0].length
<22> table_chars = [-1]*page_length
<23> for table_id, table in enumerate(tables_on_page):
<24> for span in table.spans:
<25> # replace all table spans with "table_id" in table_chars array
<26> for i in range(span.length):
<27> idx = span.offset - page_offset + i
<28> if idx >=0 and idx < page_length:
<29> table_chars[idx] = table_id
<30>
<31> </s>
|
===========below chunk 0===========
# module: scripts.prepdocs
def get_document_text(filename):
# offset: 1
page_text = ""
added_tables = set()
for idx, table_id in enumerate(table_chars):
if table_id == -1:
page_text += form_recognizer_results.content[page_offset + idx]
elif not table_id in added_tables:
page_text += table_to_html(tables_on_page[table_id])
added_tables.add(table_id)
page_text += " "
page_map.append((page_num, offset, page_text))
offset += len(page_text)
return page_map
===========unchanged ref 0===========
at: scripts.prepdocs
table_to_html(table)
args = parser.parse_args()
formrecognizer_creds = default_creds if args.formrecognizerkey is None else AzureKeyCredential(args.formrecognizerkey)
===========changed ref 0===========
# module: scripts.prepdocs
def remove_blobs(filename):
if args.verbose: print(f"Removing blobs for '{filename or '<all>'}'")
blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
blob_container = blob_service.get_container_client(args.container)
if blob_container.exists():
+ if filename is None:
- if filename == None:
blobs = blob_container.list_blob_names()
else:
prefix = os.path.splitext(os.path.basename(filename))[0]
blobs = filter(lambda b: re.match(f"{prefix}-\d+\.pdf", b), blob_container.list_blob_names(name_starts_with=os.path.splitext(os.path.basename(prefix))[0]))
for b in blobs:
if args.verbose: print(f"\tRemoving blob {b}")
blob_container.delete_blob(b)
|
|
scripts.prepdocs/split_text
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
07a97f3a74b75e6e938cec471fabd1fa7a9111f1
|
Run ruff, add to precommit (#491)
|
<5>:<add> num_pages = len(page_map)
<del> l = len(page_map)
<6>:<add> for i in range(num_pages - 1):
<del> for i in range(l - 1):
<9>:<add> return num_pages - 1
<del> return l - 1
|
# module: scripts.prepdocs
def split_text(page_map):
<0> SENTENCE_ENDINGS = [".", "!", "?"]
<1> WORDS_BREAKS = [",", ";", ":", " ", "(", ")", "[", "]", "{", "}", "\t", "\n"]
<2> if args.verbose: print(f"Splitting '{filename}' into sections")
<3>
<4> def find_page(offset):
<5> l = len(page_map)
<6> for i in range(l - 1):
<7> if offset >= page_map[i][1] and offset < page_map[i + 1][1]:
<8> return i
<9> return l - 1
<10>
<11> all_text = "".join(p[2] for p in page_map)
<12> length = len(all_text)
<13> start = 0
<14> end = length
<15> while start + SECTION_OVERLAP < length:
<16> last_word = -1
<17> end = start + MAX_SECTION_LENGTH
<18>
<19> if end > length:
<20> end = length
<21> else:
<22> # Try to find the end of the sentence
<23> while end < length and (end - start - MAX_SECTION_LENGTH) < SENTENCE_SEARCH_LIMIT and all_text[end] not in SENTENCE_ENDINGS:
<24> if all_text[end] in WORDS_BREAKS:
<25> last_word = end
<26> end += 1
<27> if end < length and all_text[end] not in SENTENCE_ENDINGS and last_word > 0:
<28> end = last_word # Fall back to at least keeping a whole word
<29> if end < length:
<30> end += 1
<31>
<32> # Try to find the start of the sentence or at least a whole word boundary
<33> last_word = -1
<34> while start > 0 and start > end - MAX_SECTION_LENGTH - 2 * SENTENCE_SEARCH_LIMIT and all_text[start] not in SENTENCE_ENDINGS:
<35> if all_text[start] in WORDS_BREAKS:
<36> last_word =</s>
|
===========below chunk 0===========
# module: scripts.prepdocs
def split_text(page_map):
# offset: 1
start -= 1
if all_text[start] not in SENTENCE_ENDINGS and last_word > 0:
start = last_word
if start > 0:
start += 1
section_text = all_text[start:end]
yield (section_text, find_page(start))
last_table_start = section_text.rfind("<table")
if (last_table_start > 2 * SENTENCE_SEARCH_LIMIT and last_table_start > section_text.rfind("</table")):
# If the section ends with an unclosed table, we need to start the next section with the table.
# If table starts inside SENTENCE_SEARCH_LIMIT, we ignore it, as that will cause an infinite loop for tables longer than MAX_SECTION_LENGTH
# If last table starts inside SECTION_OVERLAP, keep overlapping
if args.verbose: print(f"Section ends with unclosed table, starting next section with the table at page {find_page(start)} offset {start} table start {last_table_start}")
start = min(end - SECTION_OVERLAP, start + last_table_start)
else:
start = end - SECTION_OVERLAP
if start + SECTION_OVERLAP < end:
yield (all_text[start:end], find_page(start))
===========unchanged ref 0===========
at: scripts.prepdocs
MAX_SECTION_LENGTH = 1000
SENTENCE_SEARCH_LIMIT = 100
SECTION_OVERLAP = 100
args = parser.parse_args()
===========changed ref 0===========
# module: scripts.prepdocs
def remove_blobs(filename):
if args.verbose: print(f"Removing blobs for '{filename or '<all>'}'")
blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
blob_container = blob_service.get_container_client(args.container)
if blob_container.exists():
+ if filename is None:
- if filename == None:
blobs = blob_container.list_blob_names()
else:
prefix = os.path.splitext(os.path.basename(filename))[0]
blobs = filter(lambda b: re.match(f"{prefix}-\d+\.pdf", b), blob_container.list_blob_names(name_starts_with=os.path.splitext(os.path.basename(prefix))[0]))
for b in blobs:
if args.verbose: print(f"\tRemoving blob {b}")
blob_container.delete_blob(b)
===========changed ref 1===========
# module: scripts.prepdocs
def get_document_text(filename):
offset = 0
page_map = []
if args.localpdfparser:
reader = PdfReader(filename)
pages = reader.pages
for page_num, p in enumerate(pages):
page_text = p.extract_text()
page_map.append((page_num, offset, page_text))
offset += len(page_text)
else:
if args.verbose: print(f"Extracting text from '{filename}' using Azure Form Recognizer")
form_recognizer_client = DocumentAnalysisClient(endpoint=f"https://{args.formrecognizerservice}.cognitiveservices.azure.com/", credential=formrecognizer_creds, headers={"x-ms-useragent": "azure-search-chat-demo/1.0.0"})
with open(filename, "rb") as f:
poller = form_recognizer_client.begin_analyze_document("prebuilt-layout", document = f)
form_recognizer_results = poller.result()
for page_num, page in enumerate(form_recognizer_results.pages):
tables_on_page = [table for table in form_recognizer_results.tables if table.bounding_regions[0].page_number == page_num + 1]
# mark all positions of the table spans in the page
page_offset = page.spans[0].offset
page_length = page.spans[0].length
table_chars = [-1]*page_length
for table_id, table in enumerate(tables_on_page):
for span in table.spans:
# replace all table spans with "table_id" in table_chars array
for i in range(span.length):
idx = span.offset - page_offset + i
if idx >=0 and idx < page_length:
table_chars[idx] = table_id
# build page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
</s>
===========changed ref 2===========
# module: scripts.prepdocs
def get_document_text(filename):
# offset: 1
<s> page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
for idx, table_id in enumerate(table_chars):
if table_id == -1:
page_text += form_recognizer_results.content[page_offset + idx]
+ elif table_id not in added_tables:
- elif not table_id in added_tables:
page_text += table_to_html(tables_on_page[table_id])
added_tables.add(table_id)
page_text += " "
page_map.append((page_num, offset, page_text))
offset += len(page_text)
return page_map
|
scripts.prepdocs/before_retry_sleep
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
07a97f3a74b75e6e938cec471fabd1fa7a9111f1
|
Run ruff, add to precommit (#491)
|
<0>:<add> if args.verbose: print("Rate limited on the OpenAI embeddings API, sleeping before retrying...")
<del> if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
|
# module: scripts.prepdocs
def before_retry_sleep(retry_state):
<0> if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
<1>
|
===========unchanged ref 0===========
at: scripts.prepdocs
args = parser.parse_args()
===========changed ref 0===========
# module: scripts.prepdocs
def remove_blobs(filename):
if args.verbose: print(f"Removing blobs for '{filename or '<all>'}'")
blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
blob_container = blob_service.get_container_client(args.container)
if blob_container.exists():
+ if filename is None:
- if filename == None:
blobs = blob_container.list_blob_names()
else:
prefix = os.path.splitext(os.path.basename(filename))[0]
blobs = filter(lambda b: re.match(f"{prefix}-\d+\.pdf", b), blob_container.list_blob_names(name_starts_with=os.path.splitext(os.path.basename(prefix))[0]))
for b in blobs:
if args.verbose: print(f"\tRemoving blob {b}")
blob_container.delete_blob(b)
===========changed ref 1===========
# module: scripts.prepdocs
def get_document_text(filename):
offset = 0
page_map = []
if args.localpdfparser:
reader = PdfReader(filename)
pages = reader.pages
for page_num, p in enumerate(pages):
page_text = p.extract_text()
page_map.append((page_num, offset, page_text))
offset += len(page_text)
else:
if args.verbose: print(f"Extracting text from '{filename}' using Azure Form Recognizer")
form_recognizer_client = DocumentAnalysisClient(endpoint=f"https://{args.formrecognizerservice}.cognitiveservices.azure.com/", credential=formrecognizer_creds, headers={"x-ms-useragent": "azure-search-chat-demo/1.0.0"})
with open(filename, "rb") as f:
poller = form_recognizer_client.begin_analyze_document("prebuilt-layout", document = f)
form_recognizer_results = poller.result()
for page_num, page in enumerate(form_recognizer_results.pages):
tables_on_page = [table for table in form_recognizer_results.tables if table.bounding_regions[0].page_number == page_num + 1]
# mark all positions of the table spans in the page
page_offset = page.spans[0].offset
page_length = page.spans[0].length
table_chars = [-1]*page_length
for table_id, table in enumerate(tables_on_page):
for span in table.spans:
# replace all table spans with "table_id" in table_chars array
for i in range(span.length):
idx = span.offset - page_offset + i
if idx >=0 and idx < page_length:
table_chars[idx] = table_id
# build page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
</s>
===========changed ref 2===========
# module: scripts.prepdocs
def get_document_text(filename):
# offset: 1
<s> page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
for idx, table_id in enumerate(table_chars):
if table_id == -1:
page_text += form_recognizer_results.content[page_offset + idx]
+ elif table_id not in added_tables:
- elif not table_id in added_tables:
page_text += table_to_html(tables_on_page[table_id])
added_tables.add(table_id)
page_text += " "
page_map.append((page_num, offset, page_text))
offset += len(page_text)
return page_map
===========changed ref 3===========
# module: scripts.prepdocs
def split_text(page_map):
SENTENCE_ENDINGS = [".", "!", "?"]
WORDS_BREAKS = [",", ";", ":", " ", "(", ")", "[", "]", "{", "}", "\t", "\n"]
if args.verbose: print(f"Splitting '{filename}' into sections")
def find_page(offset):
+ num_pages = len(page_map)
- l = len(page_map)
+ for i in range(num_pages - 1):
- for i in range(l - 1):
if offset >= page_map[i][1] and offset < page_map[i + 1][1]:
return i
+ return num_pages - 1
- return l - 1
all_text = "".join(p[2] for p in page_map)
length = len(all_text)
start = 0
end = length
while start + SECTION_OVERLAP < length:
last_word = -1
end = start + MAX_SECTION_LENGTH
if end > length:
end = length
else:
# Try to find the end of the sentence
while end < length and (end - start - MAX_SECTION_LENGTH) < SENTENCE_SEARCH_LIMIT and all_text[end] not in SENTENCE_ENDINGS:
if all_text[end] in WORDS_BREAKS:
last_word = end
end += 1
if end < length and all_text[end] not in SENTENCE_ENDINGS and last_word > 0:
end = last_word # Fall back to at least keeping a whole word
if end < length:
end += 1
# Try to find the start of the sentence or at least a whole word boundary
last_word = -1
while start > 0 and start > end - MAX_SECTION_LENGTH - 2 * SENTENCE_SEARCH_LIMIT and all_text[start] not in SENTENCE_ENDINGS:
if all_text[start] in WORDS_BREAKS:
last_</s>
===========changed ref 4===========
# module: scripts.prepdocs
def split_text(page_map):
# offset: 1
<s>] not in SENTENCE_ENDINGS:
if all_text[start] in WORDS_BREAKS:
last_word = start
start -= 1
if all_text[start] not in SENTENCE_ENDINGS and last_word > 0:
start = last_word
if start > 0:
start += 1
section_text = all_text[start:end]
yield (section_text, find_page(start))
last_table_start = section_text.rfind("<table")
if (last_table_start > 2 * SENTENCE_SEARCH_LIMIT and last_table_start > section_text.rfind("</table")):
# If the section ends with an unclosed table, we need to start the next section with the table.
# If table starts inside SENTENCE_SEARCH_LIMIT, we ignore it, as that will cause an infinite loop for tables longer than MAX_SECTION_LENGTH
# If last table starts inside SECTION_OVERLAP, keep overlapping
if args.verbose: print(f"Section ends with unclosed table, starting next section with the table at page {find_page(start)} offset {start} table start {last_table_start}")
start = min(end - SECTION_OVERLAP, start + last_table_start)
else:
start = end - SECTION_OVERLAP
+
-
if start + SECTION_OVERLAP < end:
yield (all_text[start:end], find_page(start))
|
scripts.prepdocs/remove_from_index
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
07a97f3a74b75e6e938cec471fabd1fa7a9111f1
|
Run ruff, add to precommit (#491)
|
<5>:<add> filter = None if filename is None else f"sourcefile eq '{os.path.basename(filename)}'"
<del> filter = None if filename == None else f"sourcefile eq '{os.path.basename(filename)}'"
|
# module: scripts.prepdocs
def remove_from_index(filename):
<0> if args.verbose: print(f"Removing sections from '{filename or '<all>'}' from search index '{args.index}'")
<1> search_client = SearchClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
<2> index_name=args.index,
<3> credential=search_creds)
<4> while True:
<5> filter = None if filename == None else f"sourcefile eq '{os.path.basename(filename)}'"
<6> r = search_client.search("", filter=filter, top=1000, include_total_count=True)
<7> if r.get_count() == 0:
<8> break
<9> r = search_client.delete_documents(documents=[{ "id": d["id"] } for d in r])
<10> if args.verbose: print(f"\tRemoved {len(r)} sections from index")
<11> # It can take a few seconds for search results to reflect changes, so wait a bit
<12> time.sleep(2)
<13>
|
===========unchanged ref 0===========
at: os.path
basename(p: _PathLike[AnyStr]) -> AnyStr
basename(p: AnyStr) -> AnyStr
at: scripts.prepdocs
args = parser.parse_args()
search_creds = default_creds if args.searchkey is None else AzureKeyCredential(args.searchkey)
at: time
sleep(secs: float) -> None
===========changed ref 0===========
# module: scripts.prepdocs
def before_retry_sleep(retry_state):
+ if args.verbose: print("Rate limited on the OpenAI embeddings API, sleeping before retrying...")
- if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
===========changed ref 1===========
# module: scripts.prepdocs
def remove_blobs(filename):
if args.verbose: print(f"Removing blobs for '{filename or '<all>'}'")
blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
blob_container = blob_service.get_container_client(args.container)
if blob_container.exists():
+ if filename is None:
- if filename == None:
blobs = blob_container.list_blob_names()
else:
prefix = os.path.splitext(os.path.basename(filename))[0]
blobs = filter(lambda b: re.match(f"{prefix}-\d+\.pdf", b), blob_container.list_blob_names(name_starts_with=os.path.splitext(os.path.basename(prefix))[0]))
for b in blobs:
if args.verbose: print(f"\tRemoving blob {b}")
blob_container.delete_blob(b)
===========changed ref 2===========
# module: scripts.prepdocs
def get_document_text(filename):
offset = 0
page_map = []
if args.localpdfparser:
reader = PdfReader(filename)
pages = reader.pages
for page_num, p in enumerate(pages):
page_text = p.extract_text()
page_map.append((page_num, offset, page_text))
offset += len(page_text)
else:
if args.verbose: print(f"Extracting text from '{filename}' using Azure Form Recognizer")
form_recognizer_client = DocumentAnalysisClient(endpoint=f"https://{args.formrecognizerservice}.cognitiveservices.azure.com/", credential=formrecognizer_creds, headers={"x-ms-useragent": "azure-search-chat-demo/1.0.0"})
with open(filename, "rb") as f:
poller = form_recognizer_client.begin_analyze_document("prebuilt-layout", document = f)
form_recognizer_results = poller.result()
for page_num, page in enumerate(form_recognizer_results.pages):
tables_on_page = [table for table in form_recognizer_results.tables if table.bounding_regions[0].page_number == page_num + 1]
# mark all positions of the table spans in the page
page_offset = page.spans[0].offset
page_length = page.spans[0].length
table_chars = [-1]*page_length
for table_id, table in enumerate(tables_on_page):
for span in table.spans:
# replace all table spans with "table_id" in table_chars array
for i in range(span.length):
idx = span.offset - page_offset + i
if idx >=0 and idx < page_length:
table_chars[idx] = table_id
# build page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
</s>
===========changed ref 3===========
# module: scripts.prepdocs
def get_document_text(filename):
# offset: 1
<s> page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
for idx, table_id in enumerate(table_chars):
if table_id == -1:
page_text += form_recognizer_results.content[page_offset + idx]
+ elif table_id not in added_tables:
- elif not table_id in added_tables:
page_text += table_to_html(tables_on_page[table_id])
added_tables.add(table_id)
page_text += " "
page_map.append((page_num, offset, page_text))
offset += len(page_text)
return page_map
===========changed ref 4===========
# module: scripts.prepdocs
def split_text(page_map):
SENTENCE_ENDINGS = [".", "!", "?"]
WORDS_BREAKS = [",", ";", ":", " ", "(", ")", "[", "]", "{", "}", "\t", "\n"]
if args.verbose: print(f"Splitting '{filename}' into sections")
def find_page(offset):
+ num_pages = len(page_map)
- l = len(page_map)
+ for i in range(num_pages - 1):
- for i in range(l - 1):
if offset >= page_map[i][1] and offset < page_map[i + 1][1]:
return i
+ return num_pages - 1
- return l - 1
all_text = "".join(p[2] for p in page_map)
length = len(all_text)
start = 0
end = length
while start + SECTION_OVERLAP < length:
last_word = -1
end = start + MAX_SECTION_LENGTH
if end > length:
end = length
else:
# Try to find the end of the sentence
while end < length and (end - start - MAX_SECTION_LENGTH) < SENTENCE_SEARCH_LIMIT and all_text[end] not in SENTENCE_ENDINGS:
if all_text[end] in WORDS_BREAKS:
last_word = end
end += 1
if end < length and all_text[end] not in SENTENCE_ENDINGS and last_word > 0:
end = last_word # Fall back to at least keeping a whole word
if end < length:
end += 1
# Try to find the start of the sentence or at least a whole word boundary
last_word = -1
while start > 0 and start > end - MAX_SECTION_LENGTH - 2 * SENTENCE_SEARCH_LIMIT and all_text[start] not in SENTENCE_ENDINGS:
if all_text[start] in WORDS_BREAKS:
last_</s>
|
app.backend.langchainadapters/HtmlCallbackHandler.on_llm_start
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
07a97f3a74b75e6e938cec471fabd1fa7a9111f1
|
Run ruff, add to precommit (#491)
|
<1>:<add> self.html += "LLM prompts:<br>" + "<br>".join(ch(prompts)) + "<br>"
<del> self.html += f"LLM prompts:<br>" + "<br>".join(ch(prompts)) + "<br>";
|
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
-
def on_llm_start(
self, serialized: Dict[str, Any], prompts: List[str], **kwargs: Any
) -> None:
<0> """Print out the prompts."""
<1> self.html += f"LLM prompts:<br>" + "<br>".join(ch(prompts)) + "<br>";
<2>
|
===========unchanged ref 0===========
at: app.backend.langchainadapters.HtmlCallbackHandler
html: str = ""
at: app.backend.langchainadapters.HtmlCallbackHandler.get_and_reset_log
result = self.html
at: app.backend.langchainadapters.HtmlCallbackHandler.on_agent_action
self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_agent_finish
self.html += f"<span style='color:{color}'>{ch(finish.log)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_chain_end
self.html += "Finished chain<br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_chain_error
self.html += f"<span style='color:red'>Chain error: {ch(error)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_chain_start
self.html += f"Entering chain: {ch(class_name)}<br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_llm_error
self.html += f"<span style='color:red'>LLM error: {ch(error)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_llm_start
self.html += "LLM prompts:<br>" + "<br>".join(ch(prompts)) + "<br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_text
self.html += f"<span style='color:{color}'>{ch(text)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_tool_end
self.html += f"{ch(observation_prefix)}<br><span style='color:{color}'>{ch(output)}</span><br>{ch(llm_prefix)}<br>"
===========unchanged ref 1===========
at: app.backend.langchainadapters.HtmlCallbackHandler.on_tool_error
self.html += f"<span style='color:red'>Tool error: {ch(error)}</span><br>"
at: typing
List = _alias(list, 1, inst=False, name='List')
Dict = _alias(dict, 2, inst=False, name='Dict')
===========changed ref 0===========
# module: scripts.prepdocs
def before_retry_sleep(retry_state):
+ if args.verbose: print("Rate limited on the OpenAI embeddings API, sleeping before retrying...")
- if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
===========changed ref 1===========
# module: scripts.prepdocs
def remove_blobs(filename):
if args.verbose: print(f"Removing blobs for '{filename or '<all>'}'")
blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
blob_container = blob_service.get_container_client(args.container)
if blob_container.exists():
+ if filename is None:
- if filename == None:
blobs = blob_container.list_blob_names()
else:
prefix = os.path.splitext(os.path.basename(filename))[0]
blobs = filter(lambda b: re.match(f"{prefix}-\d+\.pdf", b), blob_container.list_blob_names(name_starts_with=os.path.splitext(os.path.basename(prefix))[0]))
for b in blobs:
if args.verbose: print(f"\tRemoving blob {b}")
blob_container.delete_blob(b)
===========changed ref 2===========
# module: scripts.prepdocs
def remove_from_index(filename):
if args.verbose: print(f"Removing sections from '{filename or '<all>'}' from search index '{args.index}'")
search_client = SearchClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
index_name=args.index,
credential=search_creds)
while True:
+ filter = None if filename is None else f"sourcefile eq '{os.path.basename(filename)}'"
- filter = None if filename == None else f"sourcefile eq '{os.path.basename(filename)}'"
r = search_client.search("", filter=filter, top=1000, include_total_count=True)
if r.get_count() == 0:
break
r = search_client.delete_documents(documents=[{ "id": d["id"] } for d in r])
if args.verbose: print(f"\tRemoved {len(r)} sections from index")
# It can take a few seconds for search results to reflect changes, so wait a bit
time.sleep(2)
===========changed ref 3===========
# module: scripts.prepdocs
def get_document_text(filename):
offset = 0
page_map = []
if args.localpdfparser:
reader = PdfReader(filename)
pages = reader.pages
for page_num, p in enumerate(pages):
page_text = p.extract_text()
page_map.append((page_num, offset, page_text))
offset += len(page_text)
else:
if args.verbose: print(f"Extracting text from '{filename}' using Azure Form Recognizer")
form_recognizer_client = DocumentAnalysisClient(endpoint=f"https://{args.formrecognizerservice}.cognitiveservices.azure.com/", credential=formrecognizer_creds, headers={"x-ms-useragent": "azure-search-chat-demo/1.0.0"})
with open(filename, "rb") as f:
poller = form_recognizer_client.begin_analyze_document("prebuilt-layout", document = f)
form_recognizer_results = poller.result()
for page_num, page in enumerate(form_recognizer_results.pages):
tables_on_page = [table for table in form_recognizer_results.tables if table.bounding_regions[0].page_number == page_num + 1]
# mark all positions of the table spans in the page
page_offset = page.spans[0].offset
page_length = page.spans[0].length
table_chars = [-1]*page_length
for table_id, table in enumerate(tables_on_page):
for span in table.spans:
# replace all table spans with "table_id" in table_chars array
for i in range(span.length):
idx = span.offset - page_offset + i
if idx >=0 and idx < page_length:
table_chars[idx] = table_id
# build page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
</s>
===========changed ref 4===========
# module: scripts.prepdocs
def get_document_text(filename):
# offset: 1
<s> page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
for idx, table_id in enumerate(table_chars):
if table_id == -1:
page_text += form_recognizer_results.content[page_offset + idx]
+ elif table_id not in added_tables:
- elif not table_id in added_tables:
page_text += table_to_html(tables_on_page[table_id])
added_tables.add(table_id)
page_text += " "
page_map.append((page_num, offset, page_text))
offset += len(page_text)
return page_map
|
app.backend.langchainadapters/HtmlCallbackHandler.on_chain_end
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
07a97f3a74b75e6e938cec471fabd1fa7a9111f1
|
Run ruff, add to precommit (#491)
|
<1>:<add> self.html += "Finished chain<br>"
<del> self.html += f"Finished chain<br>"
|
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
def on_chain_end(self, outputs: Dict[str, Any], **kwargs: Any) -> None:
<0> """Print out that we finished a chain."""
<1> self.html += f"Finished chain<br>"
<2>
|
===========unchanged ref 0===========
at: app.backend.langchainadapters
ch(text: Union[str, object]) -> str
at: app.backend.langchainadapters.HtmlCallbackHandler.get_and_reset_log
self.html = ""
at: app.backend.langchainadapters.HtmlCallbackHandler.on_agent_action
self.html += f"<span style='color:{color}'>{ch(action.log)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_agent_finish
self.html += f"<span style='color:{color}'>{ch(finish.log)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_chain_end
self.html += "Finished chain<br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_chain_error
self.html += f"<span style='color:red'>Chain error: {ch(error)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_chain_start
class_name = serialized["name"]
at: app.backend.langchainadapters.HtmlCallbackHandler.on_llm_error
self.html += f"<span style='color:red'>LLM error: {ch(error)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_llm_start
self.html += "LLM prompts:<br>" + "<br>".join(ch(prompts)) + "<br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_text
self.html += f"<span style='color:{color}'>{ch(text)}</span><br>"
at: app.backend.langchainadapters.HtmlCallbackHandler.on_tool_end
self.html += f"{ch(observation_prefix)}<br><span style='color:{color}'>{ch(output)}</span><br>{ch(llm_prefix)}<br>"
===========unchanged ref 1===========
at: app.backend.langchainadapters.HtmlCallbackHandler.on_tool_error
self.html += f"<span style='color:red'>Tool error: {ch(error)}</span><br>"
at: typing
Dict = _alias(dict, 2, inst=False, name='Dict')
===========changed ref 0===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
-
def on_llm_start(
self, serialized: Dict[str, Any], prompts: List[str], **kwargs: Any
) -> None:
"""Print out the prompts."""
+ self.html += "LLM prompts:<br>" + "<br>".join(ch(prompts)) + "<br>"
- self.html += f"LLM prompts:<br>" + "<br>".join(ch(prompts)) + "<br>";
===========changed ref 1===========
# module: scripts.prepdocs
def before_retry_sleep(retry_state):
+ if args.verbose: print("Rate limited on the OpenAI embeddings API, sleeping before retrying...")
- if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
===========changed ref 2===========
# module: scripts.prepdocs
def remove_blobs(filename):
if args.verbose: print(f"Removing blobs for '{filename or '<all>'}'")
blob_service = BlobServiceClient(account_url=f"https://{args.storageaccount}.blob.core.windows.net", credential=storage_creds)
blob_container = blob_service.get_container_client(args.container)
if blob_container.exists():
+ if filename is None:
- if filename == None:
blobs = blob_container.list_blob_names()
else:
prefix = os.path.splitext(os.path.basename(filename))[0]
blobs = filter(lambda b: re.match(f"{prefix}-\d+\.pdf", b), blob_container.list_blob_names(name_starts_with=os.path.splitext(os.path.basename(prefix))[0]))
for b in blobs:
if args.verbose: print(f"\tRemoving blob {b}")
blob_container.delete_blob(b)
===========changed ref 3===========
# module: scripts.prepdocs
def remove_from_index(filename):
if args.verbose: print(f"Removing sections from '{filename or '<all>'}' from search index '{args.index}'")
search_client = SearchClient(endpoint=f"https://{args.searchservice}.search.windows.net/",
index_name=args.index,
credential=search_creds)
while True:
+ filter = None if filename is None else f"sourcefile eq '{os.path.basename(filename)}'"
- filter = None if filename == None else f"sourcefile eq '{os.path.basename(filename)}'"
r = search_client.search("", filter=filter, top=1000, include_total_count=True)
if r.get_count() == 0:
break
r = search_client.delete_documents(documents=[{ "id": d["id"] } for d in r])
if args.verbose: print(f"\tRemoved {len(r)} sections from index")
# It can take a few seconds for search results to reflect changes, so wait a bit
time.sleep(2)
===========changed ref 4===========
# module: scripts.prepdocs
def get_document_text(filename):
offset = 0
page_map = []
if args.localpdfparser:
reader = PdfReader(filename)
pages = reader.pages
for page_num, p in enumerate(pages):
page_text = p.extract_text()
page_map.append((page_num, offset, page_text))
offset += len(page_text)
else:
if args.verbose: print(f"Extracting text from '{filename}' using Azure Form Recognizer")
form_recognizer_client = DocumentAnalysisClient(endpoint=f"https://{args.formrecognizerservice}.cognitiveservices.azure.com/", credential=formrecognizer_creds, headers={"x-ms-useragent": "azure-search-chat-demo/1.0.0"})
with open(filename, "rb") as f:
poller = form_recognizer_client.begin_analyze_document("prebuilt-layout", document = f)
form_recognizer_results = poller.result()
for page_num, page in enumerate(form_recognizer_results.pages):
tables_on_page = [table for table in form_recognizer_results.tables if table.bounding_regions[0].page_number == page_num + 1]
# mark all positions of the table spans in the page
page_offset = page.spans[0].offset
page_length = page.spans[0].length
table_chars = [-1]*page_length
for table_id, table in enumerate(tables_on_page):
for span in table.spans:
# replace all table spans with "table_id" in table_chars array
for i in range(span.length):
idx = span.offset - page_offset + i
if idx >=0 and idx < page_length:
table_chars[idx] = table_id
# build page text by replacing charcters in table spans with table html
page_text = ""
added_tables = set()
</s>
|
app.backend.approaches.retrievethenread/RetrieveThenReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
07a97f3a74b75e6e938cec471fabd1fa7a9111f1
|
Run ruff, add to precommit (#491)
|
<18>:<add> r = self.search_client.search(query_text,
<del> r = self.search_client.search(query_text,
<20>:<add> query_type=QueryType.SEMANTIC,
<del> query_type=QueryType.SEMANTIC,
<21>:<add> query_language="en-us",
<del> query_language="en-us",
<22>:<add> query_speller="lexicon",
<del> query_speller="lexicon",
<23>:<add> semantic_configuration_name="default",
<del> semantic_configuration_name="default",
<24>:<add> top=top,
<del> top=top,
<26>:<add> vector=query_vector,
<del> vector=query_vector,
<27>:<add> top_k=50 if query_vector else None,
<del> top_k=50 if query_vector else None,
|
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
def run(self, q: str, overrides: dict[str, Any]) -> Any:
<0> has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
<1> has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
<2> use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
<3> top = overrides.get("top") or 3
<4> exclude_category = overrides.get("exclude_category") or None
<5> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<6>
<7> # If retrieval mode includes vectors, compute an embedding for the query
<8> if has_vector:
<9> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=q)["data"][0]["embedding"]
<10> else:
<11> query_vector = None
<12>
<13> # Only keep the text query if the retrieval mode uses text, otherwise drop it
<14> query_text = q if has_text else None
<15>
<16> # Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
<17> if overrides.get("semantic_ranker") and has_text:
<18> r = self.search_client.search(query_text,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None,
<26> vector=query_vector,
<27> top_k=50 if query_vector else None,
<28> vector_fields="embedding" if query_vector else None)
<29> else:
</s>
|
===========below chunk 0===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
filter=filter,
top=top,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
if use_semantic_captions:
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
message_builder = MessageBuilder(overrides.get("prompt_template") or self.system_chat_template, self.chatgpt_model);
# add user question
user_content = q + "\n" + "Sources:\n {content}".format(content=content)
message_builder.append_message('user', user_content)
# Add shots/samples. This helps model to mimic response and make sure they match rules laid out in system message.
message_builder.append_message('assistant', self.answer)
message_builder.append_message('user', self.question)
messages = message_builder.messages
chat_completion = openai.ChatCompletion.create(
deployment_id=self.openai_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.3,
max_tokens=1024,
n=1)
return {"data_points": results, "answer": chat_completion.choices[0].message.content, "thoughts": f"Question:<br>{query_text}<br><br>Prompt:<br>" + '\n\</s>
===========below chunk 1===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 2
<s>, "thoughts": f"Question:<br>{query_text}<br><br>Prompt:<br>" + '\n\n'.join([str(message) for message in messages])}
===========unchanged ref 0===========
at: app.backend.approaches.retrievethenread.RetrieveThenReadApproach.__init__
self.search_client = search_client
self.embedding_deployment = embedding_deployment
at: approaches.approach.Approach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: core.messagebuilder
MessageBuilder(system_content: str, chatgpt_model: str)
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.embedding
Embedding(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.embedding.Embedding
OBJECT_NAME = "embeddings"
create(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
===========unchanged ref 1===========
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
"""
Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
top documents from search, then constructs a prompt with them, and then uses OpenAI to generate an completion
(answer) with that prompt.
"""
system_chat_template = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " + \
"Use 'you' to refer to the individual asking the questions even if they ask with 'I'. " + \
"Answer the following question using only the data provided in the sources below. " + \
"For tabular information return it as an html table. Do not return markdown format. " + \
"Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. " + \
"If you cannot answer using the sources below, say you don't know. Use below example to answer"
#shots/sample conversation
question = """
+ 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
- 'What is the deductible for the employee plan for a visit to Overlake in Bellevue?'
Sources:
info1.txt: deductibles depend on whether you are in-network or out-of-network. In-network deductibles are $500 for employee and $1000 for family. Out-of-network deductibles are $1000 for employee and $2000 for family.
info2.pdf: Overlake is in-network for the employee plan.
info3.pdf: Overlake is the name of the area that includes a park and ride near Bellevue.
info4.pdf: In-network institutions include Overlake, Swedish and others in the region
"""
answer = "In-network deductibles are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan</s>
===========changed ref 1===========
# module: app.backend.approaches.retrievethenread
class RetrieveThenReadApproach(Approach):
# offset: 1
<s> are $500 for employee and $1000 for family [info1.txt] and Overlake is in-network for the employee plan [info2.pdf][info4.pdf]."
===========changed ref 2===========
# module: app.backend.langchainadapters
class HtmlCallbackHandler (BaseCallbackHandler):
def on_chain_end(self, outputs: Dict[str, Any], **kwargs: Any) -> None:
"""Print out that we finished a chain."""
+ self.html += "Finished chain<br>"
- self.html += f"Finished chain<br>"
===========changed ref 3===========
# module: scripts.prepdocs
def before_retry_sleep(retry_state):
+ if args.verbose: print("Rate limited on the OpenAI embeddings API, sleeping before retrying...")
- if args.verbose: print(f"Rate limited on the OpenAI embeddings API, sleeping before retrying...")
|
app.backend.app/content_file
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
df48d8c23b4d0ae0ec6487cecd715a9db2943f64
|
Use an app factory pattern to enable app testing (#495)
|
<0>:<add> blob_container = current_app.config[CONFIG_BLOB_CLIENT].get_container_client(AZURE_STORAGE_CONTAINER)
|
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
+ @bp.route("/content/<path>")
- @app.route("/content/<path>")
def content_file(path):
<0> blob = blob_container.get_blob_client(path).download_blob()
<1> if not blob.properties or not blob.properties.has_key("content_settings"):
<2> abort(404)
<3> mime_type = blob.properties["content_settings"]["content_type"]
<4> if mime_type == "application/octet-stream":
<5> mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
<6> blob_file = io.BytesIO()
<7> blob.readinto(blob_file)
<8> blob_file.seek(0)
<9> return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
<10>
|
===========unchanged ref 0===========
at: app.backend.app
CONFIG_ASK_APPROACHES = "ask_approaches"
bp = Blueprint("routes", __name__, static_folder='static')
at: logging
exception(msg: Any, *args: Any, exc_info: _ExcInfoType=..., stack_info: bool=..., extra: Optional[Dict[str, Any]]=..., **kwargs: Any) -> None
===========changed ref 0===========
# module: app.backend.app
- @app.route("/", defaults={"path": "index.html"})
- @app.route("/<path:path>")
- def static_file(path):
- return app.send_static_file(path)
-
===========changed ref 1===========
# module: app.backend.app
+ @bp.route("/assets/<path:path>")
+ def assets(path):
+ return send_from_directory("static/assets", path)
+
===========changed ref 2===========
# module: app.backend.app
+ @bp.route("/favicon.ico")
+ def favicon():
+ return bp.send_static_file("favicon.ico")
+
===========changed ref 3===========
# module: app.backend.app
+ @bp.route("/")
+ def index():
+ return bp.send_static_file("index.html")
+
===========changed ref 4===========
+ # module: tests.test_app
+
+
===========changed ref 5===========
+ # module: tests.conftest
+
+
===========changed ref 6===========
+ # module: tests.conftest
+ class MockedChatApproach(ChatReadRetrieveReadApproach):
+ def __init__(self):
+ pass
+
===========changed ref 7===========
+ # module: tests.conftest
+ @pytest.fixture()
+ def client(app):
+ return app.test_client()
+
===========changed ref 8===========
+ # module: tests.conftest
+ @pytest.fixture()
+ def runner(app):
+ return app.test_cli_runner()
+
===========changed ref 9===========
+ # module: tests.conftest
+ class MockAzureCredential:
+ def get_token(self, uri):
+ return MockToken("mock_token", 9999999999)
+
===========changed ref 10===========
+ # module: tests.conftest
+ MockToken = namedtuple("MockToken", ["token", "expires_on"])
+
===========changed ref 11===========
+ # module: tests.test_app
+ def test_index(client):
+ response = client.get("/")
+ assert response.status_code == 200
+
===========changed ref 12===========
+ # module: tests.conftest
+ class MockedAskApproach(Approach):
+ def run(self, question, overrides):
+ assert question == "What is the capital of France?"
+ return {"answer": "Paris"}
+
===========changed ref 13===========
+ # module: tests.test_app
+ def test_chat_with_unknown_approach(client):
+ response = client.post("/chat", json={"approach": "test"})
+ assert response.status_code == 400
+
===========changed ref 14===========
+ # module: tests.test_app
+ def test_ask_with_unknown_approach(client):
+ response = client.post("/ask", json={"approach": "test"})
+ assert response.status_code == 400
+
===========changed ref 15===========
+ # module: tests.test_app
+ def test_chat_request_must_be_json(client):
+ response = client.post("/chat")
+ assert response.status_code == 415
+ assert response.json["error"] == "request must be json"
+
===========changed ref 16===========
+ # module: tests.test_app
+ def test_ask_request_must_be_json(client):
+ response = client.post("/ask")
+ assert response.status_code == 415
+ assert response.json["error"] == "request must be json"
+
===========changed ref 17===========
+ # module: tests.test_app
+ def test_ask_mock_approach(client):
+ response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
+ assert response.status_code == 200
+ assert response.json["answer"] == "Paris"
+
===========changed ref 18===========
+ # module: tests.test_app
+ def test_chat_mock_approach(client):
+ response = client.post(
+ "/chat",
+ json={
+ "approach": "mock",
+ "history": [{"user": "What is the capital of France?"}],
+ },
+ )
+ assert response.status_code == 200
+ assert response.json["answer"] == "Paris"
+
===========changed ref 19===========
# module: app.backend.gunicorn.conf
max_requests = 1000
max_requests_jitter = 50
log_file = "-"
bind = "0.0.0.0"
num_cpus = multiprocessing.cpu_count()
workers = (num_cpus * 2) + 1
threads = 1 if num_cpus == 1 else 2
timeout = 600
+ worker_class = "gthread"
===========changed ref 20===========
+ # module: tests.conftest
+ class MockedChatApproach(ChatReadRetrieveReadApproach):
+ def run(self, history, overrides):
+ messages = ChatReadRetrieveReadApproach.get_messages_from_history(self, ChatReadRetrieveReadApproach.query_prompt_template, "gpt-3.5-turbo", history, "Generate search query")
+ assert messages[0]["role"] == "system"
+ assert messages[1]["content"] == "Generate search query"
+ assert messages[1]["role"] == "user"
+ return {"answer": "Paris", "data_points": [], "thoughts": ""}
+
===========changed ref 21===========
+ # module: tests.conftest
+ @pytest.fixture()
+ def app():
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ _app = backend_app.create_app()
+ _app.config.update(
+ {
+ "TESTING": True,
+ backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
+ backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
+ }
+ )
+
+ yield _app
+
|
app.backend.app/ask
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
df48d8c23b4d0ae0ec6487cecd715a9db2943f64
|
Use an app factory pattern to enable app testing (#495)
|
<0>:<add> if not request.is_json:
<del> if not request.json:
<1>:<add> return jsonify({"error": "request must be json"}), 415
<del> return jsonify({"error": "request must be json"}), 400
<4>:<add> impl = current_app.config[CONFIG_ASK_APPROACHES].get(approach)
<del> impl = ask_approaches.get(approach)
|
# module: app.backend.app
+ @bp.route("/ask", methods=["POST"])
- @app.route("/ask", methods=["POST"])
def ask():
<0> if not request.json:
<1> return jsonify({"error": "request must be json"}), 400
<2> approach = request.json["approach"]
<3> try:
<4> impl = ask_approaches.get(approach)
<5> if not impl:
<6> return jsonify({"error": "unknown approach"}), 400
<7> r = impl.run(request.json["question"], request.json.get("overrides") or {})
<8> return jsonify(r)
<9> except Exception as e:
<10> logging.exception("Exception in /ask")
<11> return jsonify({"error": str(e)}), 500
<12>
|
===========unchanged ref 0===========
at: app.backend.app
CONFIG_CHAT_APPROACHES = "chat_approaches"
bp = Blueprint("routes", __name__, static_folder='static')
at: logging
exception(msg: Any, *args: Any, exc_info: _ExcInfoType=..., stack_info: bool=..., extra: Optional[Dict[str, Any]]=..., **kwargs: Any) -> None
===========changed ref 0===========
# module: app.backend.app
- @app.route("/", defaults={"path": "index.html"})
- @app.route("/<path:path>")
- def static_file(path):
- return app.send_static_file(path)
-
===========changed ref 1===========
# module: app.backend.app
+ @bp.route("/assets/<path:path>")
+ def assets(path):
+ return send_from_directory("static/assets", path)
+
===========changed ref 2===========
# module: app.backend.app
+ @bp.route("/favicon.ico")
+ def favicon():
+ return bp.send_static_file("favicon.ico")
+
===========changed ref 3===========
# module: app.backend.app
+ @bp.route("/")
+ def index():
+ return bp.send_static_file("index.html")
+
===========changed ref 4===========
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
+ @bp.route("/content/<path>")
- @app.route("/content/<path>")
def content_file(path):
+ blob_container = current_app.config[CONFIG_BLOB_CLIENT].get_container_client(AZURE_STORAGE_CONTAINER)
blob = blob_container.get_blob_client(path).download_blob()
if not blob.properties or not blob.properties.has_key("content_settings"):
abort(404)
mime_type = blob.properties["content_settings"]["content_type"]
if mime_type == "application/octet-stream":
mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
blob_file = io.BytesIO()
blob.readinto(blob_file)
blob_file.seek(0)
return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
===========changed ref 5===========
+ # module: tests.test_app
+
+
===========changed ref 6===========
+ # module: tests.conftest
+
+
===========changed ref 7===========
+ # module: tests.conftest
+ class MockedChatApproach(ChatReadRetrieveReadApproach):
+ def __init__(self):
+ pass
+
===========changed ref 8===========
+ # module: tests.conftest
+ @pytest.fixture()
+ def client(app):
+ return app.test_client()
+
===========changed ref 9===========
+ # module: tests.conftest
+ @pytest.fixture()
+ def runner(app):
+ return app.test_cli_runner()
+
===========changed ref 10===========
+ # module: tests.conftest
+ class MockAzureCredential:
+ def get_token(self, uri):
+ return MockToken("mock_token", 9999999999)
+
===========changed ref 11===========
+ # module: tests.conftest
+ MockToken = namedtuple("MockToken", ["token", "expires_on"])
+
===========changed ref 12===========
+ # module: tests.test_app
+ def test_index(client):
+ response = client.get("/")
+ assert response.status_code == 200
+
===========changed ref 13===========
+ # module: tests.conftest
+ class MockedAskApproach(Approach):
+ def run(self, question, overrides):
+ assert question == "What is the capital of France?"
+ return {"answer": "Paris"}
+
===========changed ref 14===========
+ # module: tests.test_app
+ def test_chat_with_unknown_approach(client):
+ response = client.post("/chat", json={"approach": "test"})
+ assert response.status_code == 400
+
===========changed ref 15===========
+ # module: tests.test_app
+ def test_ask_with_unknown_approach(client):
+ response = client.post("/ask", json={"approach": "test"})
+ assert response.status_code == 400
+
===========changed ref 16===========
+ # module: tests.test_app
+ def test_chat_request_must_be_json(client):
+ response = client.post("/chat")
+ assert response.status_code == 415
+ assert response.json["error"] == "request must be json"
+
===========changed ref 17===========
+ # module: tests.test_app
+ def test_ask_request_must_be_json(client):
+ response = client.post("/ask")
+ assert response.status_code == 415
+ assert response.json["error"] == "request must be json"
+
===========changed ref 18===========
+ # module: tests.test_app
+ def test_ask_mock_approach(client):
+ response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
+ assert response.status_code == 200
+ assert response.json["answer"] == "Paris"
+
===========changed ref 19===========
+ # module: tests.test_app
+ def test_chat_mock_approach(client):
+ response = client.post(
+ "/chat",
+ json={
+ "approach": "mock",
+ "history": [{"user": "What is the capital of France?"}],
+ },
+ )
+ assert response.status_code == 200
+ assert response.json["answer"] == "Paris"
+
===========changed ref 20===========
# module: app.backend.gunicorn.conf
max_requests = 1000
max_requests_jitter = 50
log_file = "-"
bind = "0.0.0.0"
num_cpus = multiprocessing.cpu_count()
workers = (num_cpus * 2) + 1
threads = 1 if num_cpus == 1 else 2
timeout = 600
+ worker_class = "gthread"
===========changed ref 21===========
+ # module: tests.conftest
+ class MockedChatApproach(ChatReadRetrieveReadApproach):
+ def run(self, history, overrides):
+ messages = ChatReadRetrieveReadApproach.get_messages_from_history(self, ChatReadRetrieveReadApproach.query_prompt_template, "gpt-3.5-turbo", history, "Generate search query")
+ assert messages[0]["role"] == "system"
+ assert messages[1]["content"] == "Generate search query"
+ assert messages[1]["role"] == "user"
+ return {"answer": "Paris", "data_points": [], "thoughts": ""}
+
===========changed ref 22===========
+ # module: tests.conftest
+ @pytest.fixture()
+ def app():
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ _app = backend_app.create_app()
+ _app.config.update(
+ {
+ "TESTING": True,
+ backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
+ backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
+ }
+ )
+
+ yield _app
+
|
app.backend.app/chat
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
df48d8c23b4d0ae0ec6487cecd715a9db2943f64
|
Use an app factory pattern to enable app testing (#495)
|
<0>:<add> if not request.is_json:
<del> if not request.json:
<1>:<add> return jsonify({"error": "request must be json"}), 415
<del> return jsonify({"error": "request must be json"}), 400
<4>:<add> impl = current_app.config[CONFIG_CHAT_APPROACHES].get(approach)
<del> impl = chat_approaches.get(approach)
|
# module: app.backend.app
+ @bp.route("/chat", methods=["POST"])
- @app.route("/chat", methods=["POST"])
def chat():
<0> if not request.json:
<1> return jsonify({"error": "request must be json"}), 400
<2> approach = request.json["approach"]
<3> try:
<4> impl = chat_approaches.get(approach)
<5> if not impl:
<6> return jsonify({"error": "unknown approach"}), 400
<7> r = impl.run(request.json["history"], request.json.get("overrides") or {})
<8> return jsonify(r)
<9> except Exception as e:
<10> logging.exception("Exception in /chat")
<11> return jsonify({"error": str(e)}), 500
<12>
|
===========unchanged ref 0===========
at: app.backend.app
CONFIG_OPENAI_TOKEN = "openai_token"
CONFIG_CREDENTIAL = "azure_credential"
at: openai
api_key = os.environ.get("OPENAI_API_KEY")
at: time
time() -> float
===========changed ref 0===========
# module: app.backend.app
- @app.route("/", defaults={"path": "index.html"})
- @app.route("/<path:path>")
- def static_file(path):
- return app.send_static_file(path)
-
===========changed ref 1===========
# module: app.backend.app
+ @bp.route("/assets/<path:path>")
+ def assets(path):
+ return send_from_directory("static/assets", path)
+
===========changed ref 2===========
# module: app.backend.app
+ @bp.route("/favicon.ico")
+ def favicon():
+ return bp.send_static_file("favicon.ico")
+
===========changed ref 3===========
# module: app.backend.app
+ @bp.route("/")
+ def index():
+ return bp.send_static_file("index.html")
+
===========changed ref 4===========
# module: app.backend.app
+ @bp.route("/ask", methods=["POST"])
- @app.route("/ask", methods=["POST"])
def ask():
+ if not request.is_json:
- if not request.json:
+ return jsonify({"error": "request must be json"}), 415
- return jsonify({"error": "request must be json"}), 400
approach = request.json["approach"]
try:
+ impl = current_app.config[CONFIG_ASK_APPROACHES].get(approach)
- impl = ask_approaches.get(approach)
if not impl:
return jsonify({"error": "unknown approach"}), 400
r = impl.run(request.json["question"], request.json.get("overrides") or {})
return jsonify(r)
except Exception as e:
logging.exception("Exception in /ask")
return jsonify({"error": str(e)}), 500
===========changed ref 5===========
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
+ @bp.route("/content/<path>")
- @app.route("/content/<path>")
def content_file(path):
+ blob_container = current_app.config[CONFIG_BLOB_CLIENT].get_container_client(AZURE_STORAGE_CONTAINER)
blob = blob_container.get_blob_client(path).download_blob()
if not blob.properties or not blob.properties.has_key("content_settings"):
abort(404)
mime_type = blob.properties["content_settings"]["content_type"]
if mime_type == "application/octet-stream":
mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
blob_file = io.BytesIO()
blob.readinto(blob_file)
blob_file.seek(0)
return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
===========changed ref 6===========
+ # module: tests.test_app
+
+
===========changed ref 7===========
+ # module: tests.conftest
+
+
===========changed ref 8===========
+ # module: tests.conftest
+ class MockedChatApproach(ChatReadRetrieveReadApproach):
+ def __init__(self):
+ pass
+
===========changed ref 9===========
+ # module: tests.conftest
+ @pytest.fixture()
+ def client(app):
+ return app.test_client()
+
===========changed ref 10===========
+ # module: tests.conftest
+ @pytest.fixture()
+ def runner(app):
+ return app.test_cli_runner()
+
===========changed ref 11===========
+ # module: tests.conftest
+ class MockAzureCredential:
+ def get_token(self, uri):
+ return MockToken("mock_token", 9999999999)
+
===========changed ref 12===========
+ # module: tests.conftest
+ MockToken = namedtuple("MockToken", ["token", "expires_on"])
+
===========changed ref 13===========
+ # module: tests.test_app
+ def test_index(client):
+ response = client.get("/")
+ assert response.status_code == 200
+
===========changed ref 14===========
+ # module: tests.conftest
+ class MockedAskApproach(Approach):
+ def run(self, question, overrides):
+ assert question == "What is the capital of France?"
+ return {"answer": "Paris"}
+
===========changed ref 15===========
+ # module: tests.test_app
+ def test_chat_with_unknown_approach(client):
+ response = client.post("/chat", json={"approach": "test"})
+ assert response.status_code == 400
+
===========changed ref 16===========
+ # module: tests.test_app
+ def test_ask_with_unknown_approach(client):
+ response = client.post("/ask", json={"approach": "test"})
+ assert response.status_code == 400
+
===========changed ref 17===========
+ # module: tests.test_app
+ def test_chat_request_must_be_json(client):
+ response = client.post("/chat")
+ assert response.status_code == 415
+ assert response.json["error"] == "request must be json"
+
===========changed ref 18===========
+ # module: tests.test_app
+ def test_ask_request_must_be_json(client):
+ response = client.post("/ask")
+ assert response.status_code == 415
+ assert response.json["error"] == "request must be json"
+
===========changed ref 19===========
+ # module: tests.test_app
+ def test_ask_mock_approach(client):
+ response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
+ assert response.status_code == 200
+ assert response.json["answer"] == "Paris"
+
===========changed ref 20===========
+ # module: tests.test_app
+ def test_chat_mock_approach(client):
+ response = client.post(
+ "/chat",
+ json={
+ "approach": "mock",
+ "history": [{"user": "What is the capital of France?"}],
+ },
+ )
+ assert response.status_code == 200
+ assert response.json["answer"] == "Paris"
+
===========changed ref 21===========
# module: app.backend.gunicorn.conf
max_requests = 1000
max_requests_jitter = 50
log_file = "-"
bind = "0.0.0.0"
num_cpus = multiprocessing.cpu_count()
workers = (num_cpus * 2) + 1
threads = 1 if num_cpus == 1 else 2
timeout = 600
+ worker_class = "gthread"
===========changed ref 22===========
+ # module: tests.conftest
+ class MockedChatApproach(ChatReadRetrieveReadApproach):
+ def run(self, history, overrides):
+ messages = ChatReadRetrieveReadApproach.get_messages_from_history(self, ChatReadRetrieveReadApproach.query_prompt_template, "gpt-3.5-turbo", history, "Generate search query")
+ assert messages[0]["role"] == "system"
+ assert messages[1]["content"] == "Generate search query"
+ assert messages[1]["role"] == "user"
+ return {"answer": "Paris", "data_points": [], "thoughts": ""}
+
|
app.backend.app/ensure_openai_token
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
df48d8c23b4d0ae0ec6487cecd715a9db2943f64
|
Use an app factory pattern to enable app testing (#495)
|
<0>:<add> openai_token = current_app.config[CONFIG_OPENAI_TOKEN]
<del> global openai_token
<2>:<add> openai_token = current_app.config[CONFIG_CREDENTIAL].get_token("https://cognitiveservices.azure.com/.default")
<del> openai_token = azure_credential.get_token("https://cognitiveservices.azure.com/.default")
<3>:<add> current_app.config[CONFIG_OPENAI_TOKEN] = openai_token
|
# module: app.backend.app
+ @bp.before_request
- @app.before_request
def ensure_openai_token():
<0> global openai_token
<1> if openai_token.expires_on < time.time() + 60:
<2> openai_token = azure_credential.get_token("https://cognitiveservices.azure.com/.default")
<3> openai.api_key = openai_token.token
<4>
|
===========unchanged ref 0===========
at: app.backend.app
AZURE_SEARCH_SERVICE = os.getenv("AZURE_SEARCH_SERVICE", "gptkb")
AZURE_SEARCH_INDEX = os.getenv("AZURE_SEARCH_INDEX", "gptkbindex")
at: app.backend.app.create_app
azure_credential = DefaultAzureCredential(exclude_shared_token_cache_credential = True)
===========changed ref 0===========
# module: app.backend.app
- @app.route("/", defaults={"path": "index.html"})
- @app.route("/<path:path>")
- def static_file(path):
- return app.send_static_file(path)
-
===========changed ref 1===========
# module: app.backend.app
+ @bp.route("/assets/<path:path>")
+ def assets(path):
+ return send_from_directory("static/assets", path)
+
===========changed ref 2===========
# module: app.backend.app
+ @bp.route("/favicon.ico")
+ def favicon():
+ return bp.send_static_file("favicon.ico")
+
===========changed ref 3===========
# module: app.backend.app
+ @bp.route("/")
+ def index():
+ return bp.send_static_file("index.html")
+
===========changed ref 4===========
# module: app.backend.app
+ @bp.route("/chat", methods=["POST"])
- @app.route("/chat", methods=["POST"])
def chat():
+ if not request.is_json:
- if not request.json:
+ return jsonify({"error": "request must be json"}), 415
- return jsonify({"error": "request must be json"}), 400
approach = request.json["approach"]
try:
+ impl = current_app.config[CONFIG_CHAT_APPROACHES].get(approach)
- impl = chat_approaches.get(approach)
if not impl:
return jsonify({"error": "unknown approach"}), 400
r = impl.run(request.json["history"], request.json.get("overrides") or {})
return jsonify(r)
except Exception as e:
logging.exception("Exception in /chat")
return jsonify({"error": str(e)}), 500
===========changed ref 5===========
# module: app.backend.app
+ @bp.route("/ask", methods=["POST"])
- @app.route("/ask", methods=["POST"])
def ask():
+ if not request.is_json:
- if not request.json:
+ return jsonify({"error": "request must be json"}), 415
- return jsonify({"error": "request must be json"}), 400
approach = request.json["approach"]
try:
+ impl = current_app.config[CONFIG_ASK_APPROACHES].get(approach)
- impl = ask_approaches.get(approach)
if not impl:
return jsonify({"error": "unknown approach"}), 400
r = impl.run(request.json["question"], request.json.get("overrides") or {})
return jsonify(r)
except Exception as e:
logging.exception("Exception in /ask")
return jsonify({"error": str(e)}), 500
===========changed ref 6===========
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
+ @bp.route("/content/<path>")
- @app.route("/content/<path>")
def content_file(path):
+ blob_container = current_app.config[CONFIG_BLOB_CLIENT].get_container_client(AZURE_STORAGE_CONTAINER)
blob = blob_container.get_blob_client(path).download_blob()
if not blob.properties or not blob.properties.has_key("content_settings"):
abort(404)
mime_type = blob.properties["content_settings"]["content_type"]
if mime_type == "application/octet-stream":
mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
blob_file = io.BytesIO()
blob.readinto(blob_file)
blob_file.seek(0)
return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
===========changed ref 7===========
# module: app.backend.app
+ def create_app():
+ app = Flask(__name__)
+
+ # Use the current user identity to authenticate with Azure OpenAI, Cognitive Search and Blob Storage (no secrets needed,
+ # just use 'az login' locally, and managed identity when deployed on Azure). If you need to use keys, use separate AzureKeyCredential instances with the
+ # keys for each service
+ # If you encounter a blocking error during a DefaultAzureCredntial resolution, you can exclude the problematic credential by using a parameter (ex. exclude_shared_token_cache_credential=True)
+ azure_credential = DefaultAzureCredential(exclude_shared_token_cache_credential = True)
+
+ # Set up clients for Cognitive Search and Storage
+ search_client = SearchClient(
+ endpoint=f"https://{AZURE_SEARCH_SERVICE}.search.windows.net",
+ index_name=AZURE_SEARCH_INDEX,
+ credential=azure_credential)
+ blob_client = BlobServiceClient(
+ account_url=f"https://{AZURE_STORAGE_ACCOUNT}.blob.core.windows.net",
+ credential=azure_credential)
+
+ # Used by the OpenAI SDK
+ openai.api_type = "azure"
+ openai.api_base = f"https://{AZURE_OPENAI_SERVICE}.openai.azure.com"
+ openai.api_version = "2023-05-15"
+
+ # Comment these two lines out if using keys, set your API key in the OPENAI_API_KEY environment variable instead
+ openai.api_type = "azure_ad"
+ openai_token = azure_credential.get_token(
+ "https://cognitiveservices.azure.com/.default"
+ )
+ openai.api_key = openai_token.token
+
+ # Store on app.config for later use inside requests
+ app.config[CONFIG_OPENAI_TOKEN] = openai_token
+ app.config[CONFIG_CREDENTIAL] = azure_credential</s>
|
app.backend.approaches.readretrieveread/ReadRetrieveReadApproach.retrieve
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<9>:<add> query_vector = (await openai.Embedding.acreate(engine=self.embedding_deployment, input=query_text))["data"][0]["embedding"]
<del> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
<15>:<add> query_text = ""
<del> query_text = None
<19>:<add> r = await self.search_client.search(query_text,
<del> r = self.search_client.search(query_text,
|
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
<0> has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
<1> has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
<2> use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
<3> top = overrides.get("top") or 3
<4> exclude_category = overrides.get("exclude_category") or None
<5> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<6>
<7> # If retrieval mode includes vectors, compute an embedding for the query
<8> if has_vector:
<9> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
<10> else:
<11> query_vector = None
<12>
<13> # Only keep the text query if the retrieval mode uses text, otherwise drop it
<14> if not has_text:
<15> query_text = None
<16>
<17> # Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
<18> if overrides.get("semantic_ranker") and has_text:
<19> r = self.search_client.search(query_text,
<20> filter=filter,
<21> query_type=QueryType.SEMANTIC,
<22> query_language="en-us",
<23> query_speller="lexicon",
<24> semantic_configuration_name="default",
<25> top = top,
<26> query_caption="extractive|highlight-false" if use_semantic_captions else None,
<27> vector=query_vector,
<28> </s>
|
===========below chunk 0===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
# offset: 1
vector_fields="embedding" if query_vector else None)
else:
r = self.search_client.search(query_text,
filter=filter,
top=top,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
if use_semantic_captions:
self.results = [doc[self.sourcepage_field] + ":" + nonewlines(" -.- ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
self.results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:250]) for doc in r]
content = "\n".join(self.results)
return content
===========unchanged ref 0===========
at: app.backend.approaches.readretrieveread.ReadRetrieveReadApproach
template_prefix = \
"You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions. " \
"Answer the question using only the data provided in the information sources below. " \
"For tabular information return it as an html table. Do not return markdown format. " \
"Each source has a name followed by colon and the actual data, quote the source name for each piece of data you use in the response. " \
"For example, if the question is \"What color is the sky?\" and one of the information sources says \"info123: the sky is blue whenever it's not cloudy\", then answer with \"The sky is blue [info123]\" " \
"It's important to strictly follow the format where the name of the source is in square brackets at the end of the sentence, and only up to the prefix before the colon (\":\"). " \
"If there are multiple sources, cite each one in their own square brackets. For example, use \"[info343][ref-76]\" and not \"[info343,ref-76]\". " \
"Never quote tool names as sources." \
"If you cannot answer using the sources below, say that you don't know. " \
"\n\nYou can access to the following tools:"
template_suffix = """
Begin!
Question: {input}
Thought: {agent_scratchpad}"""
CognitiveSearchToolDescription = "useful for searching the Microsoft employee benefits information such as healthcare plans, retirement plans, etc."
at: app.backend.approaches.readretrieveread.ReadRetrieveReadApproach.__init__
self.search_client = search_client
self.embedding_deployment = embedding_deployment
===========unchanged ref 1===========
at: openai.api_resources.embedding
Embedding(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.embedding.Embedding
OBJECT_NAME = "embeddings"
acreate(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
+ # module: app.backend.approaches
+
+
|
app.backend.approaches.readretrieveread/ReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<del> # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
<1>:<add> retrieve_results = None
<del> self.results = None
<2>:<add> async def retrieve_and_store(q: str) -> Any:
<add> nonlocal retrieve_results
<add> retrieve_results, content = await self.retrieve(q, overrides)
<add> return content
<8>:<add> func=lambda _: 'Not implemented',
<add> coroutine=retrieve_and_store,
<del> func=lambda q: self.retrieve(q, overrides),
<22>:<add> agent = ZeroShotAgent(llm_chain = chain),
<del> agent = ZeroShotAgent(llm_chain = chain, tools = tools),
<26>:<add> result = await agent_exec.arun(q)
<del> result = agent_exec.run(q)
|
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
<0> # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
<1> self.results = None
<2>
<3> # Use to capture thought process during iterations
<4> cb_handler = HtmlCallbackHandler()
<5> cb_manager = CallbackManager(handlers=[cb_handler])
<6>
<7> acs_tool = Tool(name="CognitiveSearch",
<8> func=lambda q: self.retrieve(q, overrides),
<9> description=self.CognitiveSearchToolDescription,
<10> callbacks=cb_manager)
<11> employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
<12> tools = [acs_tool, employee_tool]
<13>
<14> prompt = ZeroShotAgent.create_prompt(
<15> tools=tools,
<16> prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
<17> suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
<18> input_variables = ["input", "agent_scratchpad"])
<19> llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
<20> chain = LLMChain(llm = llm, prompt = prompt)
<21> agent_exec = AgentExecutor.from_agent_and_tools(
<22> agent = ZeroShotAgent(llm_chain = chain, tools = tools),
<23> tools = tools,
<24> verbose = True,
<25> callback_manager = cb_manager)
<26> result = agent_exec.run(q)
<27>
<28> # Remove references to tool names that might be confused with</s>
|
===========below chunk 0===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
===========unchanged ref 0===========
at: app.backend.approaches.readretrieveread
EmployeeInfoTool(employee_name: str, callbacks: Callbacks=None)
at: approaches.approach.AskApproach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: openai
api_key = os.environ.get("OPENAI_API_KEY")
===========changed ref 0===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
# If retrieval mode includes vectors, compute an embedding for the query
if has_vector:
+ query_vector = (await openai.Embedding.acreate(engine=self.embedding_deployment, input=query_text))["data"][0]["embedding"]
- query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
else:
query_vector = None
# Only keep the text query if the retrieval mode uses text, otherwise drop it
if not has_text:
+ query_text = ""
- query_text = None
# Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
if overrides.get("semantic_ranker") and has_text:
+ r = await self.search_client.search(query_text,
- r = self.search_client.search(query_text,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
</s>
===========changed ref 1===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s>language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
top = top,
query_caption="extractive|highlight-false" if use_semantic_captions else None,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
else:
+ r = await self.search_client.search(query_text,
- r = self.search_client.search(query_text,
filter=filter,
top=top,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
if use_semantic_captions:
+ results = [doc[self.sourcepage_field] + ":" + nonewlines(" -.- ".join([c.text for c in doc['@search.captions']])) async for doc in r]
- self.results = [doc[self.sourcepage_field] + ":" + nonewlines(" -.- ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
+ results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:250]) async for doc in r]
- self.results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:250]) for</s>
===========changed ref 2===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
# offset: 2
<s> r]
+ content = "\n".join(results)
- content = "\n".join(self.results)
+ return results, content
- return content
===========changed ref 3===========
+ # module: app.backend.approaches
+
+
|
app.backend.approaches.readretrieveread/EmployeeInfoTool.__init__
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<5>:<add> self.func = lambda _: 'Not implemented'
<add> self.coroutine = self.employee_info
<del> self.func = self.employee_info
|
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
<0> super().__init__(filename="data/employeeinfo.csv",
<1> key_field="name",
<2> name="Employee",
<3> description="useful for answering questions about the employee, their benefits and other personal information",
<4> callbacks=callbacks)
<5> self.func = self.employee_info
<6> self.employee_name = employee_name
<7>
|
===========unchanged ref 0===========
at: lookuptool.CsvLookupTool
data: dict[str, str] = {}
__init__(filename: Union[str, Path], key_field: str, name: str="lookup", description: str="useful to look up details given an input key as opposite to searching data with an unstructured question", callbacks: Callbacks=None)
__init__(self, filename: Union[str, Path], key_field: str, name: str="lookup", description: str="useful to look up details given an input key as opposite to searching data with an unstructured question", callbacks: Callbacks=None)
===========changed ref 0===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
===========changed ref 1===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s> agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain = chain),
- agent = ZeroShotAgent(llm_chain = chain, tools = tools),
tools = tools,
verbose = True,
callback_manager = cb_manager)
+ result = await agent_exec.arun(q)
- result = agent_exec.run(q)
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
+ return {"data_points": retrieve_results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
- return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
===========changed ref 2===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
# If retrieval mode includes vectors, compute an embedding for the query
if has_vector:
+ query_vector = (await openai.Embedding.acreate(engine=self.embedding_deployment, input=query_text))["data"][0]["embedding"]
- query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
else:
query_vector = None
# Only keep the text query if the retrieval mode uses text, otherwise drop it
if not has_text:
+ query_text = ""
- query_text = None
# Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
if overrides.get("semantic_ranker") and has_text:
+ r = await self.search_client.search(query_text,
- r = self.search_client.search(query_text,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
</s>
===========changed ref 3===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s>language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
top = top,
query_caption="extractive|highlight-false" if use_semantic_captions else None,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
else:
+ r = await self.search_client.search(query_text,
- r = self.search_client.search(query_text,
filter=filter,
top=top,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
if use_semantic_captions:
+ results = [doc[self.sourcepage_field] + ":" + nonewlines(" -.- ".join([c.text for c in doc['@search.captions']])) async for doc in r]
- self.results = [doc[self.sourcepage_field] + ":" + nonewlines(" -.- ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
+ results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:250]) async for doc in r]
- self.results = [doc[self.sourcepage_field] + ":" + nonewlines(doc[self.content_field][:250]) for</s>
|
app.backend.core.modelhelper/get_token_limit
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<2>:<add> return MODELS_2_TOKEN_LIMITS[model_id]
<del> return MODELS_2_TOKEN_LIMITS.get(model_id)
|
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
<0> if model_id not in MODELS_2_TOKEN_LIMITS:
<1> raise ValueError("Expected model gpt-35-turbo and above")
<2> return MODELS_2_TOKEN_LIMITS.get(model_id)
<3>
|
===========unchanged ref 0===========
at: app.backend.core.modelhelper
MODELS_2_TOKEN_LIMITS = {
"gpt-35-turbo": 4000,
"gpt-3.5-turbo": 4000,
"gpt-35-turbo-16k": 16000,
"gpt-3.5-turbo-16k": 16000,
"gpt-4": 8100,
"gpt-4-32k": 32000
}
===========changed ref 0===========
+ # module: app.backend.approaches
+
+
===========changed ref 1===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 2===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
===========changed ref 3===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s> agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain = chain),
- agent = ZeroShotAgent(llm_chain = chain, tools = tools),
tools = tools,
verbose = True,
callback_manager = cb_manager)
+ result = await agent_exec.arun(q)
- result = agent_exec.run(q)
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
+ return {"data_points": retrieve_results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
- return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
===========changed ref 4===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
# If retrieval mode includes vectors, compute an embedding for the query
if has_vector:
+ query_vector = (await openai.Embedding.acreate(engine=self.embedding_deployment, input=query_text))["data"][0]["embedding"]
- query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
else:
query_vector = None
# Only keep the text query if the retrieval mode uses text, otherwise drop it
if not has_text:
+ query_text = ""
- query_text = None
# Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
if overrides.get("semantic_ranker") and has_text:
+ r = await self.search_client.search(query_text,
- r = self.search_client.search(query_text,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
</s>
|
tests.conftest/client
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> # mock the DefaultAzureCredential
<add> with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
<add> mock_default_azure_credential.return_value = MockAzureCredential()
<add> quart_app = app.create_app()
<del> return app.test_client()
|
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
<0> return app.test_client()
<1>
|
===========changed ref 0===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
===========changed ref 1===========
+ # module: app.backend.approaches
+
+
===========changed ref 2===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 3===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 4===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 5===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 6===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 7===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 8===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
===========changed ref 9===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s> agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain = chain),
- agent = ZeroShotAgent(llm_chain = chain, tools = tools),
tools = tools,
verbose = True,
callback_manager = cb_manager)
+ result = await agent_exec.arun(q)
- result = agent_exec.run(q)
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
+ return {"data_points": retrieve_results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
- return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
===========changed ref 10===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
- def retrieve(self, query_text: str, overrides: dict[str, Any]) -> Any:
has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
top = overrides.get("top") or 3
exclude_category = overrides.get("exclude_category") or None
filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
# If retrieval mode includes vectors, compute an embedding for the query
if has_vector:
+ query_vector = (await openai.Embedding.acreate(engine=self.embedding_deployment, input=query_text))["data"][0]["embedding"]
- query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
else:
query_vector = None
# Only keep the text query if the retrieval mode uses text, otherwise drop it
if not has_text:
+ query_text = ""
- query_text = None
# Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
if overrides.get("semantic_ranker") and has_text:
+ r = await self.search_client.search(query_text,
- r = self.search_client.search(query_text,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
</s>
|
tests.test_app/test_index
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> response = await client.get("/")
<del> response = client.get("/")
|
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
<0> response = client.get("/")
<1> assert response.status_code == 200
<2>
|
===========changed ref 0===========
+ # module: app.backend.approaches
+
+
===========changed ref 1===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 2===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 3===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 4===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 5===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 6===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 7===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 8===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 9===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
===========changed ref 10===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
===========changed ref 11===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s> agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain = chain),
- agent = ZeroShotAgent(llm_chain = chain, tools = tools),
tools = tools,
verbose = True,
callback_manager = cb_manager)
+ result = await agent_exec.arun(q)
- result = agent_exec.run(q)
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
+ return {"data_points": retrieve_results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
- return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
|
tests.test_app/test_ask_request_must_be_json
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> response = await client.post("/ask")
<del> response = client.post("/ask")
<2>:<add> result = await response.get_json()
<add> assert result["error"] == "request must be json"
<del> assert response.json["error"] == "request must be json"
|
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
<0> response = client.post("/ask")
<1> assert response.status_code == 415
<2> assert response.json["error"] == "request must be json"
<3>
|
===========changed ref 0===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 1===========
+ # module: app.backend.approaches
+
+
===========changed ref 2===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 3===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 4===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 5===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 6===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 7===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 8===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 9===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 10===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
===========changed ref 11===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
===========changed ref 12===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s> agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain = chain),
- agent = ZeroShotAgent(llm_chain = chain, tools = tools),
tools = tools,
verbose = True,
callback_manager = cb_manager)
+ result = await agent_exec.arun(q)
- result = agent_exec.run(q)
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
+ return {"data_points": retrieve_results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
- return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
|
tests.test_app/test_ask_with_unknown_approach
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> response = await client.post("/ask", json={"approach": "test"})
<del> response = client.post("/ask", json={"approach": "test"})
|
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
<0> response = client.post("/ask", json={"approach": "test"})
<1> assert response.status_code == 400
<2>
|
===========changed ref 0===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 1===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 2===========
+ # module: app.backend.approaches
+
+
===========changed ref 3===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 4===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 5===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 6===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 7===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 8===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 9===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 10===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 11===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
===========changed ref 12===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
===========changed ref 13===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s> agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain = chain),
- agent = ZeroShotAgent(llm_chain = chain, tools = tools),
tools = tools,
verbose = True,
callback_manager = cb_manager)
+ result = await agent_exec.arun(q)
- result = agent_exec.run(q)
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
+ return {"data_points": retrieve_results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
- return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
|
tests.test_app/test_ask_mock_approach
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
<del> response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
<2>:<add> result = await response.get_json()
<add> assert result["answer"] == "Paris"
<del> assert response.json["answer"] == "Paris"
|
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
<0> response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
<1> assert response.status_code == 200
<2> assert response.json["answer"] == "Paris"
<3>
|
===========unchanged ref 0===========
at: _pytest.mark.structures
MARK_GEN = MarkGenerator(_ispytest=True)
===========changed ref 0===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 1===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 2===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 3===========
+ # module: app.backend.approaches
+
+
===========changed ref 4===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 5===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 6===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 7===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 8===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 9===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 10===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 11===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 12===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
===========changed ref 13===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
===========changed ref 14===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
<s> agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain = chain),
- agent = ZeroShotAgent(llm_chain = chain, tools = tools),
tools = tools,
verbose = True,
callback_manager = cb_manager)
+ result = await agent_exec.arun(q)
- result = agent_exec.run(q)
# Remove references to tool names that might be confused with a citation
result = result.replace("[CognitiveSearch]", "").replace("[Employee]", "")
+ return {"data_points": retrieve_results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
- return {"data_points": self.results or [], "answer": result, "thoughts": cb_handler.get_and_reset_log()}
|
tests.test_app/test_chat_request_must_be_json
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> response = await client.post("/chat")
<del> response = client.post("/chat")
<2>:<add> result = await response.get_json()
<add> assert result["error"] == "request must be json"
<del> assert response.json["error"] == "request must be json"
|
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
<0> response = client.post("/chat")
<1> assert response.status_code == 415
<2> assert response.json["error"] == "request must be json"
<3>
|
===========unchanged ref 0===========
at: _pytest.mark.structures
MARK_GEN = MarkGenerator(_ispytest=True)
===========changed ref 0===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 1===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 2===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 3===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
+ response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
- response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 4===========
+ # module: app.backend.approaches
+
+
===========changed ref 5===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 6===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 7===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 8===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 9===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 10===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 11===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 12===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 13===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
===========changed ref 14===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
|
tests.test_app/test_chat_with_unknown_approach
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> response = await client.post("/chat", json={"approach": "test"})
<del> response = client.post("/chat", json={"approach": "test"})
|
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
<0> response = client.post("/chat", json={"approach": "test"})
<1> assert response.status_code == 400
<2>
|
===========unchanged ref 0===========
at: tests.test_app.test_ask_mock_approach
result = await response.get_json()
===========changed ref 0===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 1===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 2===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 3===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 4===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
+ response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
- response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 5===========
+ # module: app.backend.approaches
+
+
===========changed ref 6===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 7===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 8===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 9===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 10===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 11===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 12===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 13===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 14===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
===========changed ref 15===========
# module: app.backend.approaches.readretrieveread
+ class ReadRetrieveReadApproach(AskApproach):
- class ReadRetrieveReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- # Not great to keep this as instance state, won't work with interleaving (e.g. if using async), but keeps the example simple
+ retrieve_results = None
- self.results = None
+ async def retrieve_and_store(q: str) -> Any:
+ nonlocal retrieve_results
+ retrieve_results, content = await self.retrieve(q, overrides)
+ return content
# Use to capture thought process during iterations
cb_handler = HtmlCallbackHandler()
cb_manager = CallbackManager(handlers=[cb_handler])
acs_tool = Tool(name="CognitiveSearch",
+ func=lambda _: 'Not implemented',
+ coroutine=retrieve_and_store,
- func=lambda q: self.retrieve(q, overrides),
description=self.CognitiveSearchToolDescription,
callbacks=cb_manager)
employee_tool = EmployeeInfoTool("Employee1", callbacks=cb_manager)
tools = [acs_tool, employee_tool]
prompt = ZeroShotAgent.create_prompt(
tools=tools,
prefix=overrides.get("prompt_template_prefix") or self.template_prefix,
suffix=overrides.get("prompt_template_suffix") or self.template_suffix,
input_variables = ["input", "agent_scratchpad"])
llm = AzureOpenAI(deployment_name=self.openai_deployment, temperature=overrides.get("temperature") or 0.3, openai_api_key=openai.api_key)
chain = LLMChain(llm = llm, prompt = prompt)
agent_exec = AgentExecutor.from_agent_and_tools(
+ agent = ZeroShotAgent(llm_chain =</s>
|
tests.test_app/test_chat_mock_approach
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> response = await client.post(
<del> response = client.post(
<8>:<add> result = await response.get_json()
<add> assert result["answer"] == "Paris"
<del> assert response.json["answer"] == "Paris"
|
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_mock_approach(client):
- def test_chat_mock_approach(client):
<0> response = client.post(
<1> "/chat",
<2> json={
<3> "approach": "mock",
<4> "history": [{"user": "What is the capital of France?"}],
<5> },
<6> )
<7> assert response.status_code == 200
<8> assert response.json["answer"] == "Paris"
<9>
|
===========unchanged ref 0===========
at: _pytest.mark.structures
MARK_GEN = MarkGenerator(_ispytest=True)
===========changed ref 0===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 1===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 2===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 3===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 4===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 5===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
+ response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
- response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 6===========
+ # module: app.backend.approaches
+
+
===========changed ref 7===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 8===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 9===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 10===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 11===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 12===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 13===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 14===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 15===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<19>:<add> chat_completion = await openai.ChatCompletion.acreate(
<del> chat_completion = openai.ChatCompletion.create(
|
# module: app.backend.approaches.chatreadretrieveread
+ class ChatReadRetrieveReadApproach(ChatApproach):
- class ChatReadRetrieveReadApproach(Approach):
+ def run(self, history: list[dict[str, str]], overrides: dict[str, Any]) -> Any:
- def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
<0> has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
<1> has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
<2> use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
<3> top = overrides.get("top") or 3
<4> exclude_category = overrides.get("exclude_category") or None
<5> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<6>
<7> user_q = 'Generate search query for: ' + history[-1]["user"]
<8>
<9> # STEP 1: Generate an optimized keyword search query based on the chat history and the last question
<10> messages = self.get_messages_from_history(
<11> self.query_prompt_template,
<12> self.chatgpt_model,
<13> history,
<14> user_q,
<15> self.query_prompt_few_shots,
<16> self.chatgpt_token_limit - len(user_q)
<17> )
<18>
<19> chat_completion = openai.ChatCompletion.create(
<20> deployment_id=self.chatgpt_deployment,
<21> model=self.chatgpt_model,
<22> messages=messages,
<23> temperature=0.0,
<24> max_tokens=32,
<25> n=1)
<26>
<27> query_text = chat_completion.choices[0].message.content
<28> if query_text.strip() == "0":
<29> query_text = history[-1]["user"] # Use the last user input if we failed to</s>
|
===========below chunk 0===========
# module: app.backend.approaches.chatreadretrieveread
+ class ChatReadRetrieveReadApproach(ChatApproach):
- class ChatReadRetrieveReadApproach(Approach):
+ def run(self, history: list[dict[str, str]], overrides: dict[str, Any]) -> Any:
- def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 1
# STEP 2: Retrieve relevant documents from the search index with the GPT optimized query
# If retrieval mode includes vectors, compute an embedding for the query
if has_vector:
query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=query_text)["data"][0]["embedding"]
else:
query_vector = None
# Only keep the text query if the retrieval mode uses text, otherwise drop it
if not has_text:
query_text = None
# Use semantic L2 reranker if requested and if retrieval mode is text or hybrid (vectors + text)
if overrides.get("semantic_ranker") and has_text:
r = self.search_client.search(query_text,
filter=filter,
query_type=QueryType.SEMANTIC,
query_language="en-us",
query_speller="lexicon",
semantic_configuration_name="default",
top=top,
query_caption="extractive|highlight-false" if use_semantic_captions else None,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
else:
r = self.search_client.search(query_text,
filter=filter,
top=top,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
if use_semantic_captions:
results = [</s>
===========below chunk 1===========
# module: app.backend.approaches.chatreadretrieveread
+ class ChatReadRetrieveReadApproach(ChatApproach):
- class ChatReadRetrieveReadApproach(Approach):
+ def run(self, history: list[dict[str, str]], overrides: dict[str, Any]) -> Any:
- def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 2
<s> vector_fields="embedding" if query_vector else None)
if use_semantic_captions:
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
follow_up_questions_prompt = self.follow_up_questions_prompt_content if overrides.get("suggest_followup_questions") else ""
# STEP 3: Generate a contextual and content specific answer using the search results and chat history
# Allow client to replace the entire prompt, or to inject into the exiting prompt using >>>
prompt_override = overrides.get("prompt_override")
if prompt_override is None:
system_message = self.system_message_chat_conversation.format(injected_prompt="", follow_up_questions_prompt=follow_up_questions_prompt)
elif prompt_override.startswith(">>>"):
system_message = self.system_message_chat_conversation.format(injected_prompt=prompt_override[3:] + "\n", follow_up_questions_prompt=follow_up_questions_prompt)
else:
system_message = prompt_override.format(follow_up_questions_prompt=follow_up_questions_prompt)
messages = self.get_messages_from_</s>
===========below chunk 2===========
# module: app.backend.approaches.chatreadretrieveread
+ class ChatReadRetrieveReadApproach(ChatApproach):
- class ChatReadRetrieveReadApproach(Approach):
+ def run(self, history: list[dict[str, str]], overrides: dict[str, Any]) -> Any:
- def run(self, history: Sequence[dict[str, str]], overrides: dict[str, Any]) -> Any:
# offset: 3
<s>
system_message + "\n\nSources:\n" + content,
self.chatgpt_model,
history,
history[-1]["user"],
max_tokens=self.chatgpt_token_limit)
chat_completion = openai.ChatCompletion.create(
deployment_id=self.chatgpt_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.7,
max_tokens=1024,
n=1)
chat_content = chat_completion.choices[0].message.content
msg_to_display = '\n\n'.join([str(message) for message in messages])
return {"data_points": results, "answer": chat_content, "thoughts": f"Searched for:<br>{query_text}<br><br>Conversations:<br>" + msg_to_display.replace('\n', '<br>')}
===========unchanged ref 0===========
at: app.backend.approaches.chatreadretrieveread.ChatReadRetrieveReadApproach
SYSTEM = "system"
USER = "user"
ASSISTANT = "assistant"
system_message_chat_conversation = """Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.
For tabular information return it as an html table. Do not return markdown format. If the question is not in English, answer in the language used in the question.
Each source has a name followed by colon and the actual information, always include the source name for each fact you use in the response. Use square brackets to reference the source, e.g. [info1.txt]. Don't combine sources, list each source separately, e.g. [info1.txt][info2.pdf].
{follow_up_questions_prompt}
{injected_prompt}
"""
follow_up_questions_prompt_content = """Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
Try not to repeat questions that have already been asked.
Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'"""
|
app.backend.approaches.chatreadretrieveread/ChatReadRetrieveReadApproach.get_messages_from_history
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<12>:<add> if bot_msg := h.get("bot"):
<del> if h.get("bot"):
<13>:<add> message_builder.append_message(self.ASSISTANT, bot_msg, index=append_index)
<del> message_builder.append_message(self.ASSISTANT, h.get('bot'), index=append_index)
<14>:<add> if user_msg := h.get("user"):
<add> message_builder.append_message(self.USER, user_msg, index=append_index)
<del> message_builder.append_message(self.USER, h.get('user'), index=append_index)
|
<s>(Approach):
+ def get_messages_from_history(self, system_prompt: str, model_id: str, history: list[dict[str, str]], user_conv: str, few_shots = [], max_tokens: int = 4096) -> list:
- def get_messages_from_history(self, system_prompt: str, model_id: str, history: Sequence[dict[str, str]], user_conv: str, few_shots = [], max_tokens: int = 4096) -> []:
<0> message_builder = MessageBuilder(system_prompt, model_id)
<1>
<2> # Add examples to show the chat what responses we want. It will try to mimic any responses and make sure they match the rules laid out in the system message.
<3> for shot in few_shots:
<4> message_builder.append_message(shot.get('role'), shot.get('content'))
<5>
<6> user_content = user_conv
<7> append_index = len(few_shots) + 1
<8>
<9> message_builder.append_message(self.USER, user_content, index=append_index)
<10>
<11> for h in reversed(history[:-1]):
<12> if h.get("bot"):
<13> message_builder.append_message(self.ASSISTANT, h.get('bot'), index=append_index)
<14> message_builder.append_message(self.USER, h.get('user'), index=append_index)
<15> if message_builder.token_length > max_tokens:
<16> break
<17>
<18> messages = message_builder.messages
<19> return messages
<20>
|
===========unchanged ref 0===========
at: core.messagebuilder
MessageBuilder(system_content: str, chatgpt_model: str)
===========changed ref 0===========
+ # module: app.backend.approaches
+
+
===========changed ref 1===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 2===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 3===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 4===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 5===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 6===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 7===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 8===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 9===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 10===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 11===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 12===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 13===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_mock_approach(client):
- def test_chat_mock_approach(client):
+ response = await client.post(
- response = client.post(
"/chat",
json={
"approach": "mock",
"history": [{"user": "What is the capital of France?"}],
},
)
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 14===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
+ response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
- response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 15===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 16===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
|
app.backend.approaches.retrievethenread/RetrieveThenReadApproach.run
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<9>:<add> query_vector = (await openai.Embedding.acreate(engine=self.embedding_deployment, input=q))["data"][0]["embedding"]
<del> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=q)["data"][0]["embedding"]
<14>:<add> query_text = q if has_text else ""
<del> query_text = q if has_text else None
<18>:<add> r = await self.search_client.search(query_text,
<del> r = self.search_client.search(query_text,
|
# module: app.backend.approaches.retrievethenread
+ class RetrieveThenReadApproach(AskApproach):
- class RetrieveThenReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
<0> has_text = overrides.get("retrieval_mode") in ["text", "hybrid", None]
<1> has_vector = overrides.get("retrieval_mode") in ["vectors", "hybrid", None]
<2> use_semantic_captions = True if overrides.get("semantic_captions") and has_text else False
<3> top = overrides.get("top") or 3
<4> exclude_category = overrides.get("exclude_category") or None
<5> filter = "category ne '{}'".format(exclude_category.replace("'", "''")) if exclude_category else None
<6>
<7> # If retrieval mode includes vectors, compute an embedding for the query
<8> if has_vector:
<9> query_vector = openai.Embedding.create(engine=self.embedding_deployment, input=q)["data"][0]["embedding"]
<10> else:
<11> query_vector = None
<12>
<13> # Only keep the text query if the retrieval mode uses text, otherwise drop it
<14> query_text = q if has_text else None
<15>
<16> # Use semantic ranker if requested and if retrieval mode is text or hybrid (vectors + text)
<17> if overrides.get("semantic_ranker") and has_text:
<18> r = self.search_client.search(query_text,
<19> filter=filter,
<20> query_type=QueryType.SEMANTIC,
<21> query_language="en-us",
<22> query_speller="lexicon",
<23> semantic_configuration_name="default",
<24> top=top,
<25> query_caption="extractive|highlight-false" if use_semantic_captions else None,
<26> vector=query_vector,
<27> top_k=50 if</s>
|
===========below chunk 0===========
# module: app.backend.approaches.retrievethenread
+ class RetrieveThenReadApproach(AskApproach):
- class RetrieveThenReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 1
vector_fields="embedding" if query_vector else None)
else:
r = self.search_client.search(query_text,
filter=filter,
top=top,
vector=query_vector,
top_k=50 if query_vector else None,
vector_fields="embedding" if query_vector else None)
if use_semantic_captions:
results = [doc[self.sourcepage_field] + ": " + nonewlines(" . ".join([c.text for c in doc['@search.captions']])) for doc in r]
else:
results = [doc[self.sourcepage_field] + ": " + nonewlines(doc[self.content_field]) for doc in r]
content = "\n".join(results)
message_builder = MessageBuilder(overrides.get("prompt_template") or self.system_chat_template, self.chatgpt_model)
# add user question
user_content = q + "\n" + f"Sources:\n {content}"
message_builder.append_message('user', user_content)
# Add shots/samples. This helps model to mimic response and make sure they match rules laid out in system message.
message_builder.append_message('assistant', self.answer)
message_builder.append_message('user', self.question)
messages = message_builder.messages
chat_completion = openai.ChatCompletion.create(
deployment_id=self.openai_deployment,
model=self.chatgpt_model,
messages=messages,
temperature=overrides.get("temperature") or 0.3,
max_tokens=1024,
n</s>
===========below chunk 1===========
# module: app.backend.approaches.retrievethenread
+ class RetrieveThenReadApproach(AskApproach):
- class RetrieveThenReadApproach(Approach):
+ def run(self, q: str, overrides: dict[str, Any]) -> Any:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
# offset: 2
<s>=messages,
temperature=overrides.get("temperature") or 0.3,
max_tokens=1024,
n=1)
return {"data_points": results, "answer": chat_completion.choices[0].message.content, "thoughts": f"Question:<br>{query_text}<br><br>Prompt:<br>" + '\n\n'.join([str(message) for message in messages])}
===========unchanged ref 0===========
at: app.backend.approaches.retrievethenread.RetrieveThenReadApproach.__init__
self.search_client = search_client
self.embedding_deployment = embedding_deployment
at: approaches.approach.AskApproach
run(self, q: str, overrides: dict[str, Any]) -> Any
at: core.messagebuilder
MessageBuilder(system_content: str, chatgpt_model: str)
at: openai.api_resources.chat_completion
ChatCompletion(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.chat_completion.ChatCompletion
engine_required = False
OBJECT_NAME = "chat.completions"
acreate(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: openai.api_resources.embedding
Embedding(engine: Optional[str]=None, *, id=None, api_key=None, api_version=None, api_type=None, organization=None, response_ms: Optional[int]=None, api_base=None, **params)
at: openai.api_resources.embedding.Embedding
OBJECT_NAME = "embeddings"
acreate(api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, /, *, api_key=None, api_base=None, api_type=None, request_id=None, api_version=None, organization=None, **params)
at: text
nonewlines(s: str) -> str
===========unchanged ref 1===========
at: typing.Mapping
get(key: _KT, default: Union[_VT_co, _T]) -> Union[_VT_co, _T]
get(key: _KT) -> Optional[_VT_co]
===========changed ref 0===========
+ # module: app.backend.approaches
+
+
===========changed ref 1===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 2===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 3===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 4===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 5===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 6===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 7===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 8===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 9===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
|
app.backend.app/index
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> return await bp.send_static_file("index.html")
<del> return bp.send_static_file("index.html")
|
# module: app.backend.app
@bp.route("/")
+ async def index():
- def index():
<0> return bp.send_static_file("index.html")
<1>
|
===========unchanged ref 0===========
at: app.backend.app
bp = Blueprint("routes", __name__, static_folder='static')
===========changed ref 0===========
+ # module: app.backend.approaches
+
+
===========changed ref 1===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 2===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 3===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 4===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 5===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 6===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 7===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 8===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 9===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 10===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 11===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 12===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 13===========
# module: app.backend.gunicorn.conf
max_requests = 1000
max_requests_jitter = 50
log_file = "-"
bind = "0.0.0.0"
+ timeout = 600
num_cpus = multiprocessing.cpu_count()
workers = (num_cpus * 2) + 1
- threads = 1 if num_cpus == 1 else 2
- timeout = 600
- worker_class = "gthread"
+ worker_class = "uvicorn.workers.UvicornWorker"
===========changed ref 14===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_mock_approach(client):
- def test_chat_mock_approach(client):
+ response = await client.post(
- response = client.post(
"/chat",
json={
"approach": "mock",
"history": [{"user": "What is the capital of France?"}],
},
)
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 15===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
+ response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
- response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 16===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 17===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
|
app.backend.app/favicon
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> return await bp.send_static_file("favicon.ico")
<del> return bp.send_static_file("favicon.ico")
|
# module: app.backend.app
@bp.route("/favicon.ico")
+ async def favicon():
- def favicon():
<0> return bp.send_static_file("favicon.ico")
<1>
|
===========unchanged ref 0===========
at: app.backend.app
bp = Blueprint("routes", __name__, static_folder='static')
===========changed ref 0===========
# module: app.backend.app
@bp.route("/")
+ async def index():
- def index():
+ return await bp.send_static_file("index.html")
- return bp.send_static_file("index.html")
===========changed ref 1===========
+ # module: app.backend.approaches
+
+
===========changed ref 2===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 3===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 4===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 5===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 6===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 7===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 8===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 9===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 10===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 11===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 12===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 13===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 14===========
# module: app.backend.gunicorn.conf
max_requests = 1000
max_requests_jitter = 50
log_file = "-"
bind = "0.0.0.0"
+ timeout = 600
num_cpus = multiprocessing.cpu_count()
workers = (num_cpus * 2) + 1
- threads = 1 if num_cpus == 1 else 2
- timeout = 600
- worker_class = "gthread"
+ worker_class = "uvicorn.workers.UvicornWorker"
===========changed ref 15===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_mock_approach(client):
- def test_chat_mock_approach(client):
+ response = await client.post(
- response = client.post(
"/chat",
json={
"approach": "mock",
"history": [{"user": "What is the capital of France?"}],
},
)
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 16===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
+ response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
- response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 17===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 18===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
|
app.backend.app/assets
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<0>:<add> return await send_from_directory("static/assets", path)
<del> return send_from_directory("static/assets", path)
|
# module: app.backend.app
@bp.route("/assets/<path:path>")
+ async def assets(path):
- def assets(path):
<0> return send_from_directory("static/assets", path)
<1>
|
===========unchanged ref 0===========
at: app.backend.app
bp = Blueprint("routes", __name__, static_folder='static')
===========changed ref 0===========
# module: app.backend.app
@bp.route("/favicon.ico")
+ async def favicon():
- def favicon():
+ return await bp.send_static_file("favicon.ico")
- return bp.send_static_file("favicon.ico")
===========changed ref 1===========
# module: app.backend.app
@bp.route("/")
+ async def index():
- def index():
+ return await bp.send_static_file("index.html")
- return bp.send_static_file("index.html")
===========changed ref 2===========
+ # module: app.backend.approaches
+
+
===========changed ref 3===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 4===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 5===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 6===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 7===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 8===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 9===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 10===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 11===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 12===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 13===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 14===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 15===========
# module: app.backend.gunicorn.conf
max_requests = 1000
max_requests_jitter = 50
log_file = "-"
bind = "0.0.0.0"
+ timeout = 600
num_cpus = multiprocessing.cpu_count()
workers = (num_cpus * 2) + 1
- threads = 1 if num_cpus == 1 else 2
- timeout = 600
- worker_class = "gthread"
+ worker_class = "uvicorn.workers.UvicornWorker"
===========changed ref 16===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_mock_approach(client):
- def test_chat_mock_approach(client):
+ response = await client.post(
- response = client.post(
"/chat",
json={
"approach": "mock",
"history": [{"user": "What is the capital of France?"}],
},
)
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 17===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
+ response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
- response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 18===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
===========changed ref 19===========
# module: tests.conftest
- @pytest.fixture()
- def app():
- # mock the DefaultAzureCredential
- with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
- mock_default_azure_credential.return_value = MockAzureCredential()
- _app = backend_app.create_app()
- _app.config.update(
- {
- "TESTING": True,
- backend_app.CONFIG_ASK_APPROACHES: {"mock": MockedAskApproach()},
- backend_app.CONFIG_CHAT_APPROACHES: {"mock": MockedChatApproach()},
- }
- )
-
- yield _app
-
|
app.backend.app/content_file
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<1>:<add> blob = await blob_container.get_blob_client(path).download_blob()
<del> blob = blob_container.get_blob_client(path).download_blob()
<8>:<add> await blob.readinto(blob_file)
<del> blob.readinto(blob_file)
<10>:<add> return await send_file(blob_file, mimetype=mime_type, as_attachment=False, attachment_filename=path)
<del> return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
|
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
@bp.route("/content/<path>")
+ async def content_file(path):
- def content_file(path):
<0> blob_container = current_app.config[CONFIG_BLOB_CLIENT].get_container_client(AZURE_STORAGE_CONTAINER)
<1> blob = blob_container.get_blob_client(path).download_blob()
<2> if not blob.properties or not blob.properties.has_key("content_settings"):
<3> abort(404)
<4> mime_type = blob.properties["content_settings"]["content_type"]
<5> if mime_type == "application/octet-stream":
<6> mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
<7> blob_file = io.BytesIO()
<8> blob.readinto(blob_file)
<9> blob_file.seek(0)
<10> return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
<11>
|
===========unchanged ref 0===========
at: app.backend.app
AZURE_STORAGE_CONTAINER = os.getenv("AZURE_STORAGE_CONTAINER", "content")
CONFIG_BLOB_CLIENT = "blob_client"
bp = Blueprint("routes", __name__, static_folder='static')
at: io
BytesIO(initial_bytes: bytes=...)
at: io.BytesIO
seek(self, offset: int, whence: int=..., /) -> int
at: mimetypes
guess_type(url: Union[Text, PathLike[str]], strict: bool=...) -> Tuple[Optional[str], Optional[str]]
===========changed ref 0===========
# module: app.backend.app
@bp.route("/assets/<path:path>")
+ async def assets(path):
- def assets(path):
+ return await send_from_directory("static/assets", path)
- return send_from_directory("static/assets", path)
===========changed ref 1===========
# module: app.backend.app
@bp.route("/favicon.ico")
+ async def favicon():
- def favicon():
+ return await bp.send_static_file("favicon.ico")
- return bp.send_static_file("favicon.ico")
===========changed ref 2===========
# module: app.backend.app
@bp.route("/")
+ async def index():
- def index():
+ return await bp.send_static_file("index.html")
- return bp.send_static_file("index.html")
===========changed ref 3===========
+ # module: app.backend.approaches
+
+
===========changed ref 4===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 5===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 6===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 7===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 8===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 9===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 10===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 11===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 12===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 13===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 14===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 15===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 16===========
# module: app.backend.gunicorn.conf
max_requests = 1000
max_requests_jitter = 50
log_file = "-"
bind = "0.0.0.0"
+ timeout = 600
num_cpus = multiprocessing.cpu_count()
workers = (num_cpus * 2) + 1
- threads = 1 if num_cpus == 1 else 2
- timeout = 600
- worker_class = "gthread"
+ worker_class = "uvicorn.workers.UvicornWorker"
===========changed ref 17===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_mock_approach(client):
- def test_chat_mock_approach(client):
+ response = await client.post(
- response = client.post(
"/chat",
json={
"approach": "mock",
"history": [{"user": "What is the capital of France?"}],
},
)
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 18===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_mock_approach(client):
- def test_ask_mock_approach(client):
+ response = await client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
- response = client.post("/ask", json={"approach": "mock", "question": "What is the capital of France?"})
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
===========changed ref 19===========
# module: app.backend.approaches.readretrieveread
class EmployeeInfoTool(CsvLookupTool):
def __init__(self, employee_name: str, callbacks: Callbacks = None):
super().__init__(filename="data/employeeinfo.csv",
key_field="name",
name="Employee",
description="useful for answering questions about the employee, their benefits and other personal information",
callbacks=callbacks)
+ self.func = lambda _: 'Not implemented'
+ self.coroutine = self.employee_info
- self.func = self.employee_info
self.employee_name = employee_name
|
app.backend.app/ask
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<2>:<add> request_json = await request.get_json()
<add> approach = request_json["approach"]
<del> approach = request.json["approach"]
<7>:<add> r = await impl.run(request_json["question"], request_json.get("overrides") or {})
<del> r = impl.run(request.json["question"], request.json.get("overrides") or {})
|
# module: app.backend.app
@bp.route("/ask", methods=["POST"])
+ async def ask():
- def ask():
<0> if not request.is_json:
<1> return jsonify({"error": "request must be json"}), 415
<2> approach = request.json["approach"]
<3> try:
<4> impl = current_app.config[CONFIG_ASK_APPROACHES].get(approach)
<5> if not impl:
<6> return jsonify({"error": "unknown approach"}), 400
<7> r = impl.run(request.json["question"], request.json.get("overrides") or {})
<8> return jsonify(r)
<9> except Exception as e:
<10> logging.exception("Exception in /ask")
<11> return jsonify({"error": str(e)}), 500
<12>
|
===========unchanged ref 0===========
at: app.backend.app
CONFIG_ASK_APPROACHES = "ask_approaches"
bp = Blueprint("routes", __name__, static_folder='static')
at: logging
exception(msg: Any, *args: Any, exc_info: _ExcInfoType=..., stack_info: bool=..., extra: Optional[Dict[str, Any]]=..., **kwargs: Any) -> None
===========changed ref 0===========
# module: app.backend.app
@bp.route("/assets/<path:path>")
+ async def assets(path):
- def assets(path):
+ return await send_from_directory("static/assets", path)
- return send_from_directory("static/assets", path)
===========changed ref 1===========
# module: app.backend.app
@bp.route("/favicon.ico")
+ async def favicon():
- def favicon():
+ return await bp.send_static_file("favicon.ico")
- return bp.send_static_file("favicon.ico")
===========changed ref 2===========
# module: app.backend.app
@bp.route("/")
+ async def index():
- def index():
+ return await bp.send_static_file("index.html")
- return bp.send_static_file("index.html")
===========changed ref 3===========
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
@bp.route("/content/<path>")
+ async def content_file(path):
- def content_file(path):
blob_container = current_app.config[CONFIG_BLOB_CLIENT].get_container_client(AZURE_STORAGE_CONTAINER)
+ blob = await blob_container.get_blob_client(path).download_blob()
- blob = blob_container.get_blob_client(path).download_blob()
if not blob.properties or not blob.properties.has_key("content_settings"):
abort(404)
mime_type = blob.properties["content_settings"]["content_type"]
if mime_type == "application/octet-stream":
mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
blob_file = io.BytesIO()
+ await blob.readinto(blob_file)
- blob.readinto(blob_file)
blob_file.seek(0)
+ return await send_file(blob_file, mimetype=mime_type, as_attachment=False, attachment_filename=path)
- return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
===========changed ref 4===========
+ # module: app.backend.approaches
+
+
===========changed ref 5===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 6===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 7===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 8===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 9===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 10===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 11===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 12===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 13===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 14===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 15===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 16===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 17===========
# module: app.backend.gunicorn.conf
max_requests = 1000
max_requests_jitter = 50
log_file = "-"
bind = "0.0.0.0"
+ timeout = 600
num_cpus = multiprocessing.cpu_count()
workers = (num_cpus * 2) + 1
- threads = 1 if num_cpus == 1 else 2
- timeout = 600
- worker_class = "gthread"
+ worker_class = "uvicorn.workers.UvicornWorker"
===========changed ref 18===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_mock_approach(client):
- def test_chat_mock_approach(client):
+ response = await client.post(
- response = client.post(
"/chat",
json={
"approach": "mock",
"history": [{"user": "What is the capital of France?"}],
},
)
assert response.status_code == 200
+ result = await response.get_json()
+ assert result["answer"] == "Paris"
- assert response.json["answer"] == "Paris"
|
app.backend.app/chat
|
Modified
|
Azure-Samples~azure-search-openai-demo
|
9da71efaebcb92ab6d73546f303e34f772088534
|
Port to Quart (#503)
|
<2>:<add> request_json = await request.get_json()
<add> approach = request_json["approach"]
<del> approach = request.json["approach"]
<7>:<add> r = await impl.run(request_json["history"], request_json.get("overrides") or {})
<del> r = impl.run(request.json["history"], request.json.get("overrides") or {})
|
# module: app.backend.app
@bp.route("/chat", methods=["POST"])
+ async def chat():
- def chat():
<0> if not request.is_json:
<1> return jsonify({"error": "request must be json"}), 415
<2> approach = request.json["approach"]
<3> try:
<4> impl = current_app.config[CONFIG_CHAT_APPROACHES].get(approach)
<5> if not impl:
<6> return jsonify({"error": "unknown approach"}), 400
<7> r = impl.run(request.json["history"], request.json.get("overrides") or {})
<8> return jsonify(r)
<9> except Exception as e:
<10> logging.exception("Exception in /chat")
<11> return jsonify({"error": str(e)}), 500
<12>
|
===========unchanged ref 0===========
at: app.backend.app
CONFIG_CHAT_APPROACHES = "chat_approaches"
bp = Blueprint("routes", __name__, static_folder='static')
at: logging
exception(msg: Any, *args: Any, exc_info: _ExcInfoType=..., stack_info: bool=..., extra: Optional[Dict[str, Any]]=..., **kwargs: Any) -> None
===========changed ref 0===========
# module: app.backend.app
@bp.route("/assets/<path:path>")
+ async def assets(path):
- def assets(path):
+ return await send_from_directory("static/assets", path)
- return send_from_directory("static/assets", path)
===========changed ref 1===========
# module: app.backend.app
@bp.route("/favicon.ico")
+ async def favicon():
- def favicon():
+ return await bp.send_static_file("favicon.ico")
- return bp.send_static_file("favicon.ico")
===========changed ref 2===========
# module: app.backend.app
@bp.route("/")
+ async def index():
- def index():
+ return await bp.send_static_file("index.html")
- return bp.send_static_file("index.html")
===========changed ref 3===========
# module: app.backend.app
@bp.route("/ask", methods=["POST"])
+ async def ask():
- def ask():
if not request.is_json:
return jsonify({"error": "request must be json"}), 415
+ request_json = await request.get_json()
+ approach = request_json["approach"]
- approach = request.json["approach"]
try:
impl = current_app.config[CONFIG_ASK_APPROACHES].get(approach)
if not impl:
return jsonify({"error": "unknown approach"}), 400
+ r = await impl.run(request_json["question"], request_json.get("overrides") or {})
- r = impl.run(request.json["question"], request.json.get("overrides") or {})
return jsonify(r)
except Exception as e:
logging.exception("Exception in /ask")
return jsonify({"error": str(e)}), 500
===========changed ref 4===========
# module: app.backend.app
# Serve content files from blob storage from within the app to keep the example self-contained.
# *** NOTE *** this assumes that the content files are public, or at least that all users of the app
# can access all the files. This is also slow and memory hungry.
@bp.route("/content/<path>")
+ async def content_file(path):
- def content_file(path):
blob_container = current_app.config[CONFIG_BLOB_CLIENT].get_container_client(AZURE_STORAGE_CONTAINER)
+ blob = await blob_container.get_blob_client(path).download_blob()
- blob = blob_container.get_blob_client(path).download_blob()
if not blob.properties or not blob.properties.has_key("content_settings"):
abort(404)
mime_type = blob.properties["content_settings"]["content_type"]
if mime_type == "application/octet-stream":
mime_type = mimetypes.guess_type(path)[0] or "application/octet-stream"
blob_file = io.BytesIO()
+ await blob.readinto(blob_file)
- blob.readinto(blob_file)
blob_file.seek(0)
+ return await send_file(blob_file, mimetype=mime_type, as_attachment=False, attachment_filename=path)
- return send_file(blob_file, mimetype=mime_type, as_attachment=False, download_name=path)
===========changed ref 5===========
+ # module: app.backend.approaches
+
+
===========changed ref 6===========
# module: app.backend.approaches.approach
+ class AskApproach(ABC):
+ @abstractmethod
+ async def run(self, q: str, overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 7===========
# module: app.backend.approaches.approach
+ class ChatApproach(ABC):
+ @abstractmethod
+ async def run(self, history: list[dict], overrides: dict[str, Any]) -> Any:
+ ...
+
===========changed ref 8===========
# module: app.backend.approaches.approach
- class Approach:
- def run(self, q: str, overrides: dict[str, Any]) -> Any:
- raise NotImplementedError
-
===========changed ref 9===========
+ # module: app.backend.main
+ app = create_app()
+
===========changed ref 10===========
# module: tests.conftest
- @pytest.fixture()
- def runner(app):
- return app.test_cli_runner()
-
===========changed ref 11===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_index(client):
- def test_index(client):
+ response = await client.get("/")
- response = client.get("/")
assert response.status_code == 200
===========changed ref 12===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_with_unknown_approach(client):
- def test_chat_with_unknown_approach(client):
+ response = await client.post("/chat", json={"approach": "test"})
- response = client.post("/chat", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 13===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_with_unknown_approach(client):
- def test_ask_with_unknown_approach(client):
+ response = await client.post("/ask", json={"approach": "test"})
- response = client.post("/ask", json={"approach": "test"})
assert response.status_code == 400
===========changed ref 14===========
# module: app.backend.core.modelhelper
def get_token_limit(model_id: str) -> int:
if model_id not in MODELS_2_TOKEN_LIMITS:
raise ValueError("Expected model gpt-35-turbo and above")
+ return MODELS_2_TOKEN_LIMITS[model_id]
- return MODELS_2_TOKEN_LIMITS.get(model_id)
===========changed ref 15===========
# module: tests.conftest
+ @pytest_asyncio.fixture
- @pytest.fixture()
+ async def client():
- def client(app):
+ # mock the DefaultAzureCredential
+ with mock.patch("app.DefaultAzureCredential") as mock_default_azure_credential:
+ mock_default_azure_credential.return_value = MockAzureCredential()
+ quart_app = app.create_app()
- return app.test_client()
===========changed ref 16===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_chat_request_must_be_json(client):
- def test_chat_request_must_be_json(client):
+ response = await client.post("/chat")
- response = client.post("/chat")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
===========changed ref 17===========
# module: tests.test_app
+ @pytest.mark.asyncio
+ async def test_ask_request_must_be_json(client):
- def test_ask_request_must_be_json(client):
+ response = await client.post("/ask")
- response = client.post("/ask")
assert response.status_code == 415
+ result = await response.get_json()
+ assert result["error"] == "request must be json"
- assert response.json["error"] == "request must be json"
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.