changed to ChatAgent

agent.py

@@ -16,6 +16,7 @@ from langchain_core.tools import tool
 from langchain.tools.retriever import create_retriever_tool
 from supabase.client import Client, create_client
 
+# === Load environment ===
 load_dotenv()
 
 # === Tools ===
@@ -69,9 +70,9 @@ with open("system_prompt.txt", "r", encoding="utf-8") as f:
     system_prompt = f.read()
 sys_msg = SystemMessage(content=system_prompt)
 
-# === Embeddings and Vector Store ===
+# === Embedding and Supabase Setup ===
 embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
-supabase = create_client(os.getenv("SUPABASE_URL"), os.getenv("SUPABASE_SERVICE_KEY"))
+supabase: Client = create_client(os.environ.get("SUPABASE_URL"), os.environ.get("SUPABASE_SERVICE_KEY"))
 vector_store = SupabaseVectorStore(
     client=supabase,
     embedding=embeddings,
@@ -79,7 +80,7 @@ vector_store = SupabaseVectorStore(
     query_name="match_documents_langchain",
 )
 
-# === Tools ===
+# === Tools List ===
 tools = [multiply, add, subtract, divide, modulus, wiki_search, web_search, arvix_search]
 
 # === Graph Builder ===
@@ -96,38 +97,44 @@ def build_graph(provider: str = "groq"):
             )
         )
     else:
-        raise ValueError("Invalid provider. Choose 'google', 'groq' or 'huggingface'.")
+        raise ValueError("Invalid provider.")
 
     llm_with_tools = llm.bind_tools(tools)
 
+    def retriever(state: MessagesState):
+        query = state["messages"][-1].content
+        similar = vector_store.similarity_search(query)
+        return {"messages": [sys_msg, state["messages"][-1], HumanMessage(content=f"Reference: {similar[0].page_content}")]}
+
     def assistant(state: MessagesState):
         response = llm_with_tools.invoke(state["messages"])
-        content = response.content.strip()
-        # Extract exact match content, remove FINAL ANSWER: if present
-        if "FINAL ANSWER:" in content:
-            content = content.split("FINAL ANSWER:")[-1].strip()
-        return {"messages": [AIMessage(content=content)]}
+        return {"messages": state["messages"] + [response]}
 
-    def retriever(state: MessagesState):
-        similar_question = vector_store.similarity_search(state["messages"][0].content)
-        example_msg = HumanMessage(content=f"Reference: {similar_question[0].page_content}")
-        return {"messages": [sys_msg] + state["messages"] + [example_msg]}
+    def formatter(state: MessagesState):
+        last = state["messages"][-1].content.strip()
+        if "FINAL ANSWER:" in last:
+            answer = last.split("FINAL ANSWER:")[-1].strip()
+        else:
+            answer = last.strip()
+        return {"messages": [AIMessage(content=answer)]}
 
     builder = StateGraph(MessagesState)
     builder.add_node("retriever", retriever)
     builder.add_node("assistant", assistant)
     builder.add_node("tools", ToolNode(tools))
+    builder.add_node("formatter", formatter)
+
     builder.add_edge(START, "retriever")
     builder.add_edge("retriever", "assistant")
     builder.add_conditional_edges("assistant", tools_condition)
     builder.add_edge("tools", "assistant")
+    builder.add_edge("assistant", "formatter")
 
     return builder.compile()
 
+# === Test Entry Point ===
 if __name__ == "__main__":
-    question = "When was a picture of St. Thomas Aquinas first added to the Wikipedia page on the Principle of double effect?"
     graph = build_graph("groq")
-    messages = [HumanMessage(content=question)]
-    messages = graph.invoke({"messages": messages})
-    for m in messages["messages"]:
-        m.pretty_print()
+    messages = graph.invoke({"messages": [HumanMessage(content="What is the capital of France?")]})
+    for msg in messages["messages"]:
+        msg.pretty_print()

app.py

@@ -1,39 +1,27 @@
-""" Basic Agent Evaluation Runner"""
 import os
-import inspect
 import gradio as gr
-import requests
 import pandas as pd
-import time
+import requests
+from dotenv import load_dotenv
 from langchain_core.messages import HumanMessage
 from agent import build_graph
-import re
 
+load_dotenv()
 
-# (Keep Constants as is)
-# --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# --- Basic Agent Definition ---
-# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
-
-
 cached_answers = []
 
-def parse_final_answer(raw: str) -> str:
-    raw = raw.strip()
-    if "FINAL ANSWER:" in raw:
-        return raw.split("FINAL ANSWER:")[-1].strip()
-    return raw.split("Final Answer:")[-1].strip() if "Final Answer:" in raw else raw
-
-class BasicAgent:
+class ChatAgent:
     def __init__(self):
+        print("ChatAgent initialized with LangGraph workflow.")
         self.graph = build_graph()
 
     def __call__(self, question: str) -> str:
+        print(f"Processing question: {question[:60]}...")
         messages = [HumanMessage(content=question)]
-        output = self.graph.invoke({"messages": messages})
-        return parse_final_answer(output['messages'][-1].content)
+        results = self.graph.invoke({"messages": messages})
+        answer = results['messages'][-1].content.strip()
+        return answer
 
 def run_agent_only(profile: gr.OAuthProfile | None):
     global cached_answers
@@ -44,67 +32,74 @@ def run_agent_only(profile: gr.OAuthProfile | None):
         return "Please login first.", None
 
     try:
-        agent = BasicAgent()
+        agent = ChatAgent()
     except Exception as e:
         return f"Agent Init Error: {e}", None
 
     try:
-        questions = requests.get("https://agents-course-unit4-scoring.hf.space/questions", timeout=15).json()
+        response = requests.get(f"{DEFAULT_API_URL}/questions", timeout=15)
+        questions_data = response.json()
     except Exception as e:
         return f"Error fetching questions: {e}", None
 
-    with open("system_prompt.txt", "r") as f:
-        system_prompt = f.read().strip()
-
-    for item in questions:
+    for item in questions_data:
         task_id = item.get("task_id")
         question = item.get("question")
         file_name = item.get("file_name")
 
-        if not task_id or not question:
+        if not task_id or question is None:
             continue
 
         try:
-            user_message = question + (f"\n\nFile to use: {file_name}" if file_name else "")
-            answer = agent(system_prompt + "\n\n" + user_message)
+            user_message = question
+            if file_name:
+                user_message += f"\n\nFile to use: {file_name}"
+
+            answer = agent(user_message)
             cached_answers.append({"task_id": task_id, "submitted_answer": answer})
             results_log.append({"Task ID": task_id, "Question": question, "Submitted Answer": answer})
         except Exception as e:
             results_log.append({"Task ID": task_id, "Question": question, "Submitted Answer": f"AGENT ERROR: {e}"})
 
-    return "Agent finished. Click 'Submit Cached Answers' next.", pd.DataFrame(results_log)
+    return "Agent finished. Now click 'Submit Cached Answers'", pd.DataFrame(results_log)
 
 def submit_cached_answers(profile: gr.OAuthProfile | None):
+    global cached_answers
     if not profile or not cached_answers:
-        return "Nothing to submit. Run the agent first.", None
+        return "No cached answers to submit. Run the agent first.", None
+
+    space_id = os.getenv("SPACE_ID")
+    username = profile.username
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
 
     payload = {
-        "username": profile.username,
-        "agent_code": f"https://huggingface.co/spaces/{os.getenv('SPACE_ID')}/tree/main",
+        "username": username,
+        "agent_code": agent_code,
         "answers": cached_answers
     }
 
     try:
-        response = requests.post("https://agents-course-unit4-scoring.hf.space/submit", json=payload, timeout=60)
+        response = requests.post(f"{DEFAULT_API_URL}/submit", json=payload, timeout=60)
         result = response.json()
-        score = result.get("score", "?")
-        correct = result.get("correct_count", "?")
-        total = result.get("total_attempted", "?")
-        return f"Submission complete. Score: {score}% ({correct}/{total})", None
+        final_status = (
+            f"Submission Successful!\nUser: {result.get('username')}\n"
+            f"Score: {result.get('score', 'N/A')}% ({result.get('correct_count', '?')}/{result.get('total_attempted', '?')})"
+        )
+        return final_status, None
     except Exception as e:
         return f"Submission failed: {e}", None
 
+# --- Gradio UI ---
 with gr.Blocks() as demo:
-    gr.Markdown("""# Agent Evaluator
-1. Login with Hugging Face
-2. Run agent only
-3. Submit answers""")
-
+    gr.Markdown("# LangGraph ChatAgent Evaluation")
+    gr.Markdown("Run the agent on all tasks, then submit for scoring.")
     gr.LoginButton()
-    run_button = gr.Button("Run Agent")
-    submit_button = gr.Button("Submit Cached Answers")
-    status_box = gr.Textbox(label="Status", lines=4)
-    table = gr.DataFrame(label="Answers Log")
+
+    run_button = gr.Button("🧠 Run Agent")
+    submit_button = gr.Button("📤 Submit Answers")
+
+    status_box = gr.Textbox(label="Status", lines=3)
+    table = gr.DataFrame(label="Results", wrap=True)
 
     run_button.click(fn=run_agent_only, outputs=[status_box, table])
     submit_button.click(fn=submit_cached_answers, outputs=[status_box, table])

system_prompt.txt

@@ -1,6 +1,9 @@
-You are a helpful assistant. Think step-by-step to solve the question. Then output only the final answer as your last message. The final answer must be:
-- a number (without comma separators or symbols),
-- a string (no articles or abbreviations),
-- or a comma-separated list of such elements.
-Do not include any explanation or prefix like "Answer:", "FINAL ANSWER:", or similar.
-Return only the answer.
+You are a helpful assistant. Think step-by-step to solve the question using available tools or reasoning. When you arrive at the correct answer, respond with only the answer and nothing else.
+
+Your answer must follow these rules:
+- If it's a number, do not include commas, currency symbols, or percentage signs.
+- If it's a string, do not use articles (a, an, the) or abbreviations.
+- If it's a list, format it as a comma-separated list of numbers or strings.
+- Do not prefix your answer with "Answer:", "FINAL ANSWER:", or any other label.
+
+Your reply should contain the answer only. No explanation. No formatting. No markdown. Just the final answer.