Update app.py

app.py

@@ -2,118 +2,94 @@ import os
 import gradio as gr
 import requests
 import pandas as pd
+from typing import TypedDict, Annotated, Sequence
+from langchain_core.messages import BaseMessage
 from langchain import hub
-from langchain.agents import AgentExecutor, create_react_agent
 from langchain_community.tools import DuckDuckGoSearchRun
 from langchain_huggingface import HuggingFaceEndpoint
+from langgraph.prebuilt import create_agent_executor
 
-# --- Constants ---
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+# --- Main Application Logic ---
 
-# --- Final LangChain Agent Definition ---
-class LangChainAgent:
-    def __init__(self):
-        print("Initializing LangChainAgent...")
+# This is our agent's "memory" or state. It keeps track of the conversation.
+class AgentState(TypedDict):
+    messages: Annotated[Sequence[BaseMessage], lambda x, y: x + y]
 
-        # 1. Set up the LLM using the stable Flan-T5 model
-        self.llm = HuggingFaceEndpoint(
-            repo_id="google/flan-t5-xxl",
-            task="text-generation",
-            max_new_tokens=512,
-        )
-        print("LLM initialized with google/flan-t5-xxl.")
-
-        # 2. Define the Tools
-        self.tools = [DuckDuckGoSearchRun()]
-        print("Tools initialized.")
-
-        # 3. Get the Prompt Template
-        self.prompt = hub.pull("hwchase17/react")
-        print("Prompt template pulled.")
-
-        # 4. Create the Agent
-        self.agent = create_react_agent(self.llm, self.tools, self.prompt)
-        print("Agent created.")
+# This is the modern way to create a robust agent with LangGraph.
+# It automatically wires up the LLM, tools, and state into a graph.
+def create_langgraph_agent():
+    print("Initializing LangGraphAgent...")
+    
+    # 1. Set up the LLM (The "Brain")
+    llm = HuggingFaceEndpoint(
+        repo_id="mistralai/Mistral-7B-Instruct-v0.2",
+        task="conversational",
+        max_new_tokens=512,
+        do_sample=False,
+    )
+    print("LLM initialized.")
 
-        # 5. Create the Agent Executor with a robust error handler
-        self.agent_executor = AgentExecutor(
-            agent=self.agent,
-            tools=self.tools,
-            verbose=True,
-            # This new line gives the agent instructions on how to recover from parsing errors
-            handle_parsing_errors="Check your output and make sure it conforms to the format. Your final answer must be just the answer, with no other text.",
-            max_iterations=5,
-        )
-        print("Agent Executor created. Initialization complete.")
+    # 2. Define the Tools
+    tools = [DuckDuckGoSearchRun()]
+    print("Tools initialized.")
+    
+    # 3. Create the agent executor using LangGraph's prebuilt function
+    # This is much more robust than the previous AgentExecutor.
+    agent_executor = create_agent_executor(llm, tools, checkpointer=None)
+    print("LangGraph agent executor created.")
+    
+    return agent_executor
 
-    def __call__(self, question: str) -> str:
-        """
-        This method is executed for each question.
-        """
-        print(f"Agent received question: {question}")
-        
-        try:
-            response = self.agent_executor.invoke({"input": question})
-            final_answer = response.get("output", "Error: Could not parse final answer.")
-        except Exception as e:
-            print(f"Error during agent execution: {e}")
-            final_answer = "Error: Agent failed to execute."
-            
-        print(f"Agent returning answer: {final_answer}")
-        return str(final_answer)
+# This function runs the agent for a single question.
+def run_agent(agent_executor, question: str) -> str:
+    print(f"Agent received question: {question}")
+    try:
+        # We invoke the agent with the question in the correct message format
+        response = agent_executor.invoke({"messages": [("user", question)]})
+        # The final answer is in the last message of the output
+        final_answer = response['messages'][-1].content
+    except Exception as e:
+        print(f"Error during agent execution: {e}")
+        final_answer = f"Error: Agent failed to execute. {e}"
+    
+    print(f"Agent returning answer: {final_answer}")
+    return str(final_answer)
 
-# --- Main Application Logic (Uses our new agent) ---
 def run_and_submit_all(profile: gr.OAuthProfile | None):
     """
-    Fetches all questions, runs the LangChainAgent on them, submits all answers,
+    Fetches all questions, runs the LangGraphAgent on them, submits all answers,
     and displays the results.
     """
     space_id = os.getenv("SPACE_ID")
-
-    if profile:
-        username = f"{profile.username}"
-        print(f"User logged in: {username}")
-    else:
-        print("User not logged in.")
+    if not profile:
         return "Please Login to Hugging Face with the button.", None
-
-    api_url = DEFAULT_API_URL
-    questions_url = f"{api_url}/questions"
-    submit_url = f"{api_url}/submit"
-
-    # 1. Instantiate Agent
+    username = f"{profile.username}"
+    
     try:
-        agent = LangChainAgent()
+        agent_executor = create_langgraph_agent()
     except Exception as e:
-        print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
         
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
-
-    # 2. Fetch Questions
+    
+    questions_url = f"https://agents-course-unit4-scoring.hf.space/questions"
     try:
         response = requests.get(questions_url, timeout=20)
         response.raise_for_status()
         questions_data = response.json()
     except Exception as e:
-        return f"An unexpected error occurred fetching questions: {e}", None
+        return f"Error fetching questions: {e}", None
 
-    # 3. Run your Agent
-    results_log = []
     answers_payload = []
     for item in questions_data:
-        task_id = item.get("task_id")
-        question_text = item.get("question")
-        submitted_answer = agent(question_text)
+        task_id, question_text = item.get("task_id"), item.get("question")
+        submitted_answer = run_agent(agent_executor, question_text)
         answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
-        results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
 
-    # 4. Prepare Submission 
     submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
-    
-    # 5. Submit
+    submit_url = f"https://agents-course-unit4-scoring.hf.space/submit"
     try:
-        response = requests.post(submit_url, json=submission_data, timeout=60)
+        response = requests.post(submit_url, json=submission_data, timeout=120) # Increased timeout
         response.raise_for_status()
         result_data = response.json()
         final_status = (
@@ -123,32 +99,18 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
             f"Message: {result_data.get('message', 'No message received.')}"
         )
-        results_df = pd.DataFrame(results_log)
-        return final_status, results_df
+        return final_status, pd.DataFrame(answers_payload)
     except Exception as e:
-        status_message = f"An unexpected error occurred during submission: {e}"
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
+        return f"Error during submission: {e}", pd.DataFrame(answers_payload)
 
 # --- Gradio Interface (Unchanged) ---
 with gr.Blocks() as demo:
-    gr.Markdown("# Agent Evaluation Runner (LangChain 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.
-        """
-    )
+    gr.Markdown("# Agent Evaluation Runner (LangGraph Version)")
     gr.LoginButton()
     run_button = gr.Button("Run Evaluation & Submit All Answers")
     status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
     results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
-
-    run_button.click(
-        fn=run_and_submit_all,
-        outputs=[status_output, results_table]
-    )
+    run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
 
 if __name__ == "__main__":
     demo.launch()