0.2 changing chat_raw to embedd instructions into the model

main.py

@@ -6,6 +6,17 @@ import os
 from langchain_community.llms import LlamaCpp
 from langchain_openai import ChatOpenAI
 
+# --- Raw model identity & rules (system prompt) ---
+KAI_SYSTEM_MESSAGE = {
+    "role": "system",
+    "content": (
+        "You are Kai, a fast, direct technical assistant. "
+        "Purpose: help with debugging, deployment, Python/FastAPI, LLM ops. "
+        "Style: concise, step-by-step when needed, include exact commands, avoid fluff."
+    )
+}
+
+
 llm = ChatOpenAI(
     base_url=os.getenv("OPENAI_API_BASE"),
     api_key=os.getenv("OPENAI_API_KEY"),
@@ -62,9 +73,31 @@ def health_check():
         "rails_config_loaded": True,
     }
 
+def call_openai_chat(messages: List[Dict], **params) -> str:
+    payload = {
+        "model": "kai-model",          # or whatever your server reports
+        "messages": messages,
+        "temperature": params.get("temperature", 0.7),
+        "max_tokens": params.get("max_tokens", 128),
+        "stream": False
+    }
+    r = requests.post(
+        f"{OPENAI_API_BASE}/chat/completions",
+        headers={"Authorization": f"Bearer {OPENAI_API_KEY}",
+                 "Content-Type": "application/json"},
+        json=payload, timeout=120,
+    )
+    r.raise_for_status()
+    return r.json()["choices"][0]["message"]["content"]
+
 @app.post("/chat_raw")
 def chat_raw(r: ChatRequest):
-    return {"text": llm.invoke(r.message)}  # same llm instance
+    messages = [
+        KAI_SYSTEM_MESSAGE,                 # << always prepended
+        {"role": "user", "content": r.message}
+    ]
+    text = call_openai_chat(messages, max_tokens=128, temperature=0.7)
+    return {"text": text}
 
 
 if __name__ == "__main__":