Update app.py

app.py

@@ -1,235 +1,126 @@
+import os
+import gradio as gr
+import requests
 import pandas as pd
-import re
-from huggingface_hub import InferenceClient
-
 
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+HF_TOKEN = os.getenv("HF_TOKEN")  # Make sure your HF read token is set in environment variables
 
-
-# =========================
-# GAIA OPTIMIZED AGENT
-# =========================
-class GAIAAgent:
-    """
-    GAIA benchmark agent – chat-only, nscale-safe, exact answers.
-
-    """
-
-    def __init__(self):
-        print("🚀 GAIAAgent initializing...")
-
-        hf_token = (
-            os.getenv("HF_TOKEN")
-            or os.getenv("HUGGING_FACE_HUB_TOKEN")
-            or os.getenv("HF_API_TOKEN")
-        )
-
-        if not hf_token:
-            raise RuntimeError("HF_TOKEN not found in Space secrets")
-
-
-
-
-        self.client = InferenceClient(token=hf_token)
-
-        # ✅ SAFE MODELS (chat-only)
-        self.model = "meta-llama/Meta-Llama-3-8B-Instruct"
-        # Alternative:
-        # self.model = "Qwen/Qwen2.5-7B-Instruct"
-
-        print(f"✅ Model loaded: {self.model}")
+# --- Gaia Agent using Qwen API ---
+class GaiaAgentQwen:
+    def __init__(self, model="Qwen/Qwen2.5-Coder-32B-Instruct"):
+        self.model = model
+        self.api_url = f"https://api-inference.huggingface.co/models/{model}"
+        self.headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+        print(f"GaiaAgentQwen initialized with model {model}")
 
     def __call__(self, question: str) -> str:
-        print(f"\nQ: {question[:120]}")
-
-
-
-
-
-
-
-
-
-
+        prompt = f"Answer the following question concisely and correctly:\n{question}"
+        payload = {"inputs": prompt, "options": {"wait_for_model": True}}
         try:
-            answer = self._chat_answer(question)
-            print(f"A: {answer}")
-            return answer
+            response = requests.post(self.api_url, headers=self.headers, json=payload, timeout=60)
+            response.raise_for_status()
+            data = response.json()
+            if isinstance(data, list) and "generated_text" in data[0]:
+                return data[0]["generated_text"]
+            else:
+                return str(data)  # fallback
         except Exception as e:
-            print(f"❌ Agent error: {e}")
-            return "Unable to determine answer"
-
-    def _chat_answer(self, question: str) -> str:
-        messages = [
-            {
-                "role": "system",
-                "content": (
-                    "You are an expert GAIA benchmark solver.\n"
-                    "Answer EXACTLY what is asked.\n"
-                    "Return ONLY the final answer.\n"
-                    "No explanations, no prefixes, no formatting."
-                )
-            },
-            {
-                "role": "user",
-                "content": question
-            }
-        ]
-
-        response = self.client.chat_completion(
-            model=self.model,
-            messages=messages,
-            max_tokens=256,
-            temperature=0.0,
-
-
-        )
-
-        if not response or not response.choices:
-            return "Unable to determine answer"
-
-        raw = response.choices[0].message.content.strip()
-        return self._clean_answer(raw)
-
-    def _clean_answer(self, text: str) -> str:
-        """
-        GAIA-safe cleaning: minimal, no hallucinated trimming.
-        """
-        text = text.strip()
-
-        # Remove common junk if model disobeys
-        bad_prefixes = [
-            "answer:",
-            "final answer:",
-            "the answer is",
-            "result:"
-        ]
-
-        for p in bad_prefixes:
-            if text.lower().startswith(p):
-                text = text[len(p):].strip()
-
-        # If multi-line, keep first meaningful line
-        if "\n" in text:
-            text = text.split("\n")[0].strip()
-
-        # GAIA prefers concise
-        if len(text.split()) > 12:
-            # keep last sentence
-            parts = re.split(r"[.!?]", text)
-            text = parts[-2].strip() if len(parts) > 1 else parts[0].strip()
-
-        return text
-
-
-
-
-
-
+            print(f"Error calling HF Inference API: {e}")
+            return f"API ERROR: {e}"
 
-# =========================
-# RUN + SUBMIT
-# =========================
+# --- Main function ---
 def run_and_submit_all(profile: gr.OAuthProfile | None):
+    api_url = DEFAULT_API_URL
+    space_id = os.getenv("SPACE_ID") or "unknown-space"
 
-    if not profile:
-        return "Please login with Hugging Face.", None
-
-    username = profile.username
-    print(f"👤 User: {username}")
-
-
-
-    questions_url = f"{DEFAULT_API_URL}/questions"
-    submit_url = f"{DEFAULT_API_URL}/submit"
-
-    agent = GAIAAgent()
+    username = profile.username if profile else "anonymous"
+    if profile:
+        print(f"User logged in: {username}")
+    else:
+        print("User not logged in.")
 
+    questions_url = f"{api_url}/questions"
+    submit_url = f"{api_url}/submit"
 
+    # Instantiate the Gaia agent
+    try:
+        agent = GaiaAgentQwen()
+    except Exception as e:
+        return f"Error initializing agent: {e}", None
 
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
+    print(f"Agent code repo: {agent_code}")
 
     # Fetch questions
-    questions = requests.get(questions_url, timeout=15).json()
-
-    answers_payload = []
+    try:
+        print(f"Fetching questions from: {questions_url}")
+        response = requests.get(questions_url, timeout=10)
+        response.raise_for_status()
+        questions_data = response.json()
+        if not questions_data:
+            return "Fetched questions list is empty or invalid format.", None
+        print(f"Fetched {len(questions_data)} questions.")
+    except Exception as e:
+        return f"Error fetching questions: {e}", None
+
+    # Run agent on questions
     results_log = []
-
-    for idx, item in enumerate(questions):
-        task_id = item["task_id"]
-        question = item["question"]
-
-
-
-
-
-        print(f"\n[{idx+1}/{len(questions)}] {task_id}")
-        answer = agent(question)
-
-        answers_payload.append({
-            "task_id": task_id,
-            "submitted_answer": answer
-        })
-
-
-
-
-        results_log.append({
-            "Task ID": task_id,
-            "Answer": answer
-
-        })
-
-    submission = {
-        "username": username,
-        "agent_code": f"https://huggingface.co/spaces/{os.getenv('SPACE_ID')}",
-        "answers": answers_payload
-    }
-
-    response = requests.post(submit_url, json=submission, timeout=60)
-    result = response.json()
-
-    status = (
-        f"🎉 Submission Successful\n\n"
-        f"Score: {result.get('score')}%\n"
-        f"Correct: {result.get('correct_count')}/{result.get('total_attempted')}"
-    )
-
-    return status, pd.DataFrame(results_log)
-
-
-
-# =========================
-# GRADIO UI
-# =========================
-with gr.Blocks(title="GAIA Agent") as demo:
-    gr.Markdown("# 🤗 GAIA Benchmark Agent (Fixed)")
-
-
-
-
-    gr.Markdown(
-        """
-        ✅ Chat-only  
-        ✅ nscale-safe  
-        ✅ GAIA-optimized  
-        
-        **Steps**
-        1. Add `HF_TOKEN` to Space secrets  
-        2. Login with Hugging Face  
-        3. Click Run  
-        """
+    answers_payload = []
+    print(f"Running agent on {len(questions_data)} questions...")
+    for item in questions_data:
+        task_id = item.get("task_id")
+        question_text = item.get("question", "")
+        if not task_id or not question_text:
+            print(f"Skipping invalid question: {item}")
+            continue
+        try:
+            answer = agent(question_text)
+            answers_payload.append({"task_id": task_id, "submitted_answer": answer})
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": answer})
+        except Exception as e:
+            print(f"Error running agent on task {task_id}: {e}")
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
+
+    if not answers_payload:
+        return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
+
+    # Submit answers
+    submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
+    try:
+        print(f"Submitting {len(answers_payload)} answers for user '{username}'...")
+        response = requests.post(submit_url, json=submission_data, timeout=20)
+        response.raise_for_status()
+        submission_result = response.json()
+        print(f"Submission result: {submission_result}")
+        return "Submission completed successfully!", pd.DataFrame(results_log)
+    except Exception as e:
+        return f"Error submitting answers: {e}", pd.DataFrame(results_log)
+
+# --- Build Gradio Interface ---
+with gr.Blocks() as demo:
+    gr.Markdown("# Gaia Agent Evaluation Runner")
+    gr.Markdown("""
+    **Instructions:**
+    1. Clone this space, then modify the code to define your agent's logic.
+    2. Log in to your Hugging Face account using the button below.
+    3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see results.
+
+    **Note:** Using the HF API can take a few seconds per question.
+    """)
+    login_btn = 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,
+        inputs=[login_btn],
+        outputs=[status_output, results_table]
     )
 
-    gr.LoginButton()
-    run_btn = gr.Button("🚀 Run Evaluation", variant="primary")
-
-    status = gr.Textbox(label="Status", lines=6)
-    table = gr.DataFrame(label="Results")
-
-    run_btn.click(run_and_submit_all, outputs=[status, table])
-
-
-
 if __name__ == "__main__":
-    demo.launch(debug=True)
+    print("\n" + "-"*30 + " App Starting " + "-"*30)
+    print("Launching Gradio Interface for Gaia Agent Evaluation...")
+    demo.launch(debug=True, share=False)