Update handler.py

handler.py

@@ -37,15 +37,23 @@ class EndpointHandler:
         """
         Preprocess input data before inference
         """
-        inputs = data.pop("inputs", data)
+        # HF Inference API format: 'inputs' is the text, 'parameters' contains the config
+        # Handle both direct access and standardized HF format
+        if isinstance(data, dict) and "inputs" in data:
+            # Standard HF format
+            text = data["inputs"]
+            parameters = data.get("parameters", {})
+        else:
+            # Direct access (fallback)
+            text = data
+            parameters = {}
         
         # Extract parameters from request
-        text = inputs.get("text", "")
-        voice = inputs.get("voice", "tara")
-        temperature = float(inputs.get("temperature", 0.6))
-        top_p = float(inputs.get("top_p", 0.95))
-        max_new_tokens = int(inputs.get("max_new_tokens", 1200))
-        repetition_penalty = float(inputs.get("repetition_penalty", 1.1))
+        voice = parameters.get("voice", "tara")
+        temperature = float(parameters.get("temperature", 0.6))
+        top_p = float(parameters.get("top_p", 0.95))
+        max_new_tokens = int(parameters.get("max_new_tokens", 1200))
+        repetition_penalty = float(parameters.get("repetition_penalty", 1.1))
         
         # Format prompt with voice
         prompt = f"{voice}: {text}"
@@ -193,7 +201,19 @@ class EndpointHandler:
         """
         Main entry point for the handler
         """
-        preprocessed_inputs = self.preprocess(data)
-        model_outputs = self.inference(preprocessed_inputs)
-        response = self.postprocess(model_outputs)
-        return response
+        try:
+            logger.info(f"Received request: {type(data)}")
+            
+            # Check if we need to handle the health check route
+            if data == "ping" or data == {"inputs": "ping"}:
+                return {"status": "ok"}
+                
+            preprocessed_inputs = self.preprocess(data)
+            model_outputs = self.inference(preprocessed_inputs)
+            response = self.postprocess(model_outputs)
+            return response
+        except Exception as e:
+            logger.error(f"Error processing request: {str(e)}")
+            import traceback
+            logger.error(traceback.format_exc())
+            return {"error": str(e)}