Update app.py

app.py

@@ -1,4 +1,4 @@
-""" Enhanced Multi-LLM Agent Evaluation Runner with Agno Integration"""
+""" Enhanced Multi-LLM Agent Evaluation Runner - CORRECTED VERSION"""
 import os
 import gradio as gr
 import requests
@@ -11,56 +11,70 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
 # --- Enhanced Agent Definition ---
 class EnhancedMultiLLMAgent:
-    """A multi-provider LangGraph agent with Agno-style reasoning capabilities."""
+    """A multi-provider LangGraph agent with proper response handling."""
     def __init__(self):
-        print("Enhanced Multi-LLM Agent with Agno Integration initialized.")
+        print("Enhanced Multi-LLM Agent initialized.")
         try:
             self.graph = build_graph(provider="groq")
-            print("Enhanced Multi-LLM Graph built successfully.")
+            print("Multi-LLM Graph built successfully.")
         except Exception as e:
             print(f"Error building graph: {e}")
             self.graph = None
 
     def __call__(self, question: str) -> str:
-        print(f"Agent received question (first 50 chars): {question[:50]}...")
+        print(f"Agent received question: {question[:100]}...")
         
         if self.graph is None:
             return "Error: Agent not properly initialized"
         
-        # CRITICAL FIX: Always pass the complete state expected by the graph
+        # Create complete state structure
         state = {
             "messages": [HumanMessage(content=question)],
-            "query": question,  # This was the critical missing field
+            "query": question,  # Critical: this must match the question
             "agent_type": "",
             "final_answer": "",
             "perf": {},
-            "agno_resp": "",
-            "tools_used": [],
-            "reasoning": "",
-            "confidence": ""
+            "agno_resp": ""
         }
-        # CRITICAL FIX: Always provide the required config with thread_id
+        # Always provide the required config with thread_id
         config = {"configurable": {"thread_id": f"eval_{hash(question)}"}}
         
         try:
             result = self.graph.invoke(state, config)
             
-            # Handle different response formats
+            # CORRECTED: Proper response extraction
             if isinstance(result, dict):
-                if 'messages' in result and result['messages']:
-                    answer = result['messages'][-1].content
-                elif 'final_answer' in result:
+                # First try to get final_answer from the state
+                if 'final_answer' in result and result['final_answer']:
                     answer = result['final_answer']
+                # Fallback to messages if final_answer is empty
+                elif 'messages' in result and result['messages']:
+                    last_message = result['messages'][-1]
+                    if hasattr(last_message, 'content'):
+                        answer = last_message.content
+                    else:
+                        answer = str(last_message)
                 else:
                     answer = str(result)
             else:
                 answer = str(result)
             
+            # Clean the answer
+            answer = answer.strip()
+            
+            # CRITICAL FIX: Ensure we don't return the question as answer
+            if answer == question or answer.startswith(question):
+                return "Information not available"
+            
             # Extract final answer if present
             if "FINAL ANSWER:" in answer:
-                return answer.split("FINAL ANSWER:")[-1].strip()
-            else:
-                return answer.strip()
+                answer = answer.split("FINAL ANSWER:")[-1].strip()
+            
+            # Additional validation
+            if not answer or len(answer.strip()) == 0:
+                return "No answer generated"
+                
+            return answer
                 
         except Exception as e:
             error_msg = f"Error: {str(e)}"
@@ -68,10 +82,7 @@ class EnhancedMultiLLMAgent:
             return error_msg
 
 def run_and_submit_all(profile: gr.OAuthProfile | None):
-    """
-    Fetches all questions, runs the Enhanced Multi-LLM Agent on them, 
-    submits all answers, and displays the results.
-    """
+    """Fetch questions, run agent, and submit answers."""
     space_id = os.getenv("SPACE_ID")
     
     if profile:
@@ -114,7 +125,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
     # 3. Run your Agent
     results_log = []
     answers_payload = []
-    print(f"Running Enhanced Multi-LLM agent with Agno integration on {len(questions_data)} questions...")
+    print(f"Running Enhanced Multi-LLM agent on {len(questions_data)} questions...")
     
     for i, item in enumerate(questions_data):
         task_id = item.get("task_id")
@@ -128,6 +139,11 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         
         try:
             submitted_answer = agent(question_text)
+            
+            # Additional validation to prevent question repetition
+            if submitted_answer == question_text or submitted_answer.startswith(question_text):
+                submitted_answer = "Information not available"
+            
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             results_log.append({
                 "Task ID": task_id, 
@@ -175,36 +191,26 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         results_df = pd.DataFrame(results_log)
         return status_message, results_df
 
-# --- Build Gradio Interface using Blocks ---
+# --- Build Gradio Interface ---
 with gr.Blocks() as demo:
-    gr.Markdown("# Enhanced Multi-LLM Agent with Agno Integration")
+    gr.Markdown("# Enhanced Multi-LLM Agent - CORRECTED VERSION")
     gr.Markdown(
         """
         **Instructions:**
         1. Log in to your Hugging Face account using the button below.
         2. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
         
-        **Enhanced Agent Features:**
-        - **Multi-LLM Support**: Groq (Llama-3 8B/70B, DeepSeek), Google Gemini, NVIDIA NIM
-        - **Agno Integration**: Systematic reasoning with step-by-step analysis
-        - **Intelligent Routing**: Automatically selects best provider based on query complexity
-        - **Enhanced Tools**: Mathematical operations, web search, Wikipedia integration
-        - **Question-Answering**: Optimized for evaluation tasks with proper formatting
-        - **Error Handling**: Robust fallback mechanisms and comprehensive logging
-        
-        **Routing Examples:**
-        - Standard: "What is the capital of France?" → Llama-3 8B
-        - Complex: "Analyze quantum computing principles" → Llama-3 70B
-        - Search: "Find information about Mercedes Sosa" → Search-Enhanced
-        - Agno: "agno llama-70: Systematic analysis of AI ethics" → Agno Llama-3 70B
-        - Provider-specific: "google: Explain machine learning" → Google Gemini
+        **FIXES APPLIED:**
+        - ✅ Proper response extraction from graph state
+        - ✅ Prevention of question repetition as answer
+        - ✅ Enhanced prompt engineering for better responses
+        - ✅ Improved error handling and validation
+        - ✅ Search-enhanced processing for information retrieval
         """
     )
 
     gr.LoginButton()
-
     run_button = gr.Button("Run Evaluation & Submit All Answers", variant="primary")
-
     status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
     results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
 
@@ -214,5 +220,5 @@ with gr.Blocks() as demo:
     )
 
 if __name__ == "__main__":
-    print("\n" + "-"*30 + " Enhanced Multi-LLM Agent with Agno Starting " + "-"*30)
+    print("\n" + "-"*30 + " Enhanced Multi-LLM Agent CORRECTED Starting " + "-"*30)
     demo.launch(debug=True, share=False)