Enhance automatic validation process in agent.py. Updated response handling to provide context and next steps when guidelines are not found, ensuring professionalism. Modified validation function to run silently in the background, logging results for backend analysis without displaying them to the user. Improved error handling and logging for validation failures.

core/agent.py

@@ -119,7 +119,14 @@ You will be responding to practicing medical professionals so adjust your answer
 - When citing text:
   * Specify the section or subsection heading
   * Indicate if it's from a bullet point, paragraph, or other format
-- If the answer is not found in the retrieved guidelines, respond: "I do not know."
+
+- If the answer is not found in the retrieved guidelines, provide a helpful response that:
+  * Acknowledges the limitation: "Based on the available medical guidelines in my knowledge base, I could not find specific information about [topic]."
+  * Suggests alternatives: "You may want to:
+    - Rephrase your question with more specific clinical details
+    - Specify a particular guideline provider (NCCN, ASCO, ESMO, NICE)
+    - Consult the latest published guidelines directly for emerging topics"
+  * Maintains professionalism: Never simply say "I don't know" - always provide context and next steps
 - Never speculate or provide information not present in the guidelines.
 - Always respond in English.
 
@@ -235,6 +242,8 @@ def _should_validate_response(user_input: str, response: str) -> bool:
         "sorry,",
         "i don't know",
         "i do not know",
+        "could not find specific information",
+        "not found in the retrieved guidelines",
         "validation report",
         "evaluation scores"
     ]
@@ -271,16 +280,17 @@ def _should_validate_response(user_input: str, response: str) -> bool:
     return any(indicator in response_lower for indicator in medical_indicators)
 
 
-def _perform_automatic_validation(user_input: str, response: str) -> str:
+def _perform_automatic_validation(user_input: str, response: str) -> None:
     """
-    Perform automatic validation and append results to response.
+    Perform automatic validation in the background without displaying results to user.
+    Validation results are logged and saved to GitHub repository for backend analysis.
     
     Args:
         user_input: The user's input
         response: The agent's response
         
     Returns:
-        str: Response with validation results appended
+        None: Validation runs silently in background
     """
     try:
         # Import here to avoid circular imports
@@ -289,52 +299,24 @@ def _perform_automatic_validation(user_input: str, response: str) -> str:
         # Check if we have the necessary context for validation
         if not _last_question or not _last_documents:
             logger.info("Skipping validation: insufficient context")
-            return response
+            return
         
         # Perform validation using the original user input instead of tool query
         evaluation = validate_medical_answer(user_input, _last_documents, response)
         
-        # Format validation results
+        # Log validation results to backend only (not shown to user)
         report = evaluation.get("validation_report", {})
+        logger.info(f"Background validation completed - Interaction ID: {evaluation.get('interaction_id', 'N/A')}")
+        logger.info(f"Validation scores - Overall: {report.get('Overall_Rating', 'N/A')}/100, "
+                   f"Accuracy: {report.get('Accuracy_Rating', 'N/A')}/100, "
+                   f"Coherence: {report.get('Coherence_Rating', 'N/A')}/100, "
+                   f"Relevance: {report.get('Relevance_Rating', 'N/A')}/100")
         
-        validation_summary = f"""
-
----
-
-## 🔍 **AUTOMATIC VALIDATION REPORT**
-
-**Overall Score:** {report.get('Overall_Rating', 'N/A')}/100
-
-**Key Metrics:**
-
-**Accuracy:** {report.get('Accuracy_Rating', 'N/A')}/100
-{report.get('Accuracy_Comment', 'No comment available')}
-
-**Coherence:** {report.get('Coherence_Rating', 'N/A')}/100
-{report.get('Coherence_Comment', 'No comment available')}
-
-**Relevance:** {report.get('Relevance_Rating', 'N/A')}/100
-{report.get('Relevance_Comment', 'No comment available')}
-
-**Completeness:** {report.get('Completeness_Rating', 'N/A')}/100
-{report.get('Completeness_Comment', 'No comment available')}
-
-**Citations:** {report.get('Citations_Attribution_Rating', 'N/A')}/100
-{report.get('Citations_Attribution_Comment', 'No comment available')}
-
-**Length:** {report.get('Length_Rating', 'N/A')}/100
-{report.get('Length_Comment', 'No comment available')}
-
-**Assessment:** {report.get('Final_Summary_and_Improvement_Plan', 'No assessment available')}
-
-*Validation ID: {evaluation.get('interaction_id', 'N/A')} | Saved to evaluation_results.json*
-"""
-        
-        return response + validation_summary
+        # Validation is automatically saved to GitHub by validate_medical_answer function
+        # No need to return anything - results are stored in backend only
         
     except Exception as e:
-        logger.error(f"Automatic validation failed: {e}")
-        return response
+        logger.error(f"Background validation failed: {e}")
 
 
 # ============================================================================
@@ -447,22 +429,15 @@ async def run_agent_streaming(user_input: str, session_id: str = "default", max_
             if not response["output"] or not response["output"].strip():
                 raise ValidationError("Empty response from agent")
             
-            # Perform automatic validation if appropriate
+            # Perform automatic validation in background (hidden from user)
             base_response = response["output"]
             if _should_validate_response(user_input, base_response):
-                logger.info("Performing automatic validation for streaming response...")
+                logger.info("Performing background validation for streaming response...")
                 try:
-                    validation_content = _perform_automatic_validation(user_input, base_response)
-                    # Extract just the validation part (everything after the original response)
-                    if len(validation_content) > len(base_response):
-                        validation_part = validation_content[len(base_response):]
-                        # Stream the validation part
-                        validation_words = validation_part.split(' ')
-                        for word in validation_words:
-                            yield word + ' '
-                            await asyncio.sleep(0.02)
+                    # Run validation silently - results saved to backend/GitHub only
+                    _perform_automatic_validation(user_input, base_response)
                 except Exception as e:
-                    logger.error(f"Streaming validation failed: {e}")
+                    logger.error(f"Background validation failed: {e}")
             
             # Save conversation context to memory
             memory.save_context(
@@ -710,11 +685,15 @@ async def run_agent(user_input: str, session_id: str = "default", max_retries: i
 
             logger.info(f"Successfully processed user input: {user_input[:50]}...")
             
-            # Perform automatic validation if appropriate
+            # Perform automatic validation in background (hidden from user)
             final_response = response["output"]
             if _should_validate_response(user_input, final_response):
-                logger.info("Performing automatic validation...")
-                final_response = _perform_automatic_validation(user_input, final_response)
+                logger.info("Performing background validation...")
+                try:
+                    # Run validation silently - results saved to backend/GitHub only
+                    _perform_automatic_validation(user_input, final_response)
+                except Exception as e:
+                    logger.error(f"Background validation failed: {e}")
             
             return final_response