Update app.py

app.py

@@ -1,116 +1,37 @@
 import os
 import json
 import requests
-import openai
 import traceback
 import gradio as gr
 import pandas as pd
 from typing import Optional
 from dotenv import load_dotenv
+from transformers import pipeline
 
 # ────────────────  Load Environment  ────────────────
 load_dotenv()
-openai.api_key = os.getenv("OPENAI_API_KEY")
+hf_token = os.getenv("HF_TOKEN")
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
-# ────────────────  Utility: Python Executor  ────────────────
-def python_exec(code: str) -> str:
-    try:
-        safe_globals = {"__builtins__": {}}
-        local_vars = {}
-        exec(code, safe_globals, local_vars)
-        if "result" in local_vars:
-            return str(local_vars["result"])
-        else:
-            return "No variable named 'result' was produced."
-    except Exception:
-        return "Error during Python execution:\n" + traceback.format_exc()
-
-# ────────────────  Function Schema  ────────────────
-function_definitions = [
-    {
-        "name": "python_exec",
-        "description": "Execute Python code and return the value of `result` (or printed output).",
-        "parameters": {
-            "type": "object",
-            "properties": {
-                "code": {
-                    "type": "string",
-                    "description": "The Python code to execute. Must assign the final answer to a variable named `result`."
-                }
-            },
-            "required": ["code"]
-        }
-    }
-]
+# ────────────────  Load Zephyr Model  ────────────────
+try:
+    pipe = pipeline("text-generation", model="HuggingFaceH4/zephyr-7b-alpha", token=hf_token)
+except Exception as e:
+    raise RuntimeError(f"Model loading failed: {e}")
 
 # ────────────────  Agent Class  ────────────────
-class BasicAgent:
+class ZephyrAgent:
     def __init__(self):
-        if not openai.api_key:
-            raise RuntimeError("Please set OPENAI_API_KEY in your .env file")
-        print("✅ BasicAgent initialized with GPT-4o-mini.")
+        print("✅ ZephyrAgent initialized.")
 
     def __call__(self, question: str) -> str:
         try:
-            first_resp = openai.chat.completions.create(
-                model="gpt-4o-mini",
-                messages=[
-                    {
-                        "role": "system",
-                        "content": "You are an assistant that can solve math/programming questions by calling python_exec when needed."
-                    },
-                    {"role": "user", "content": question}
-                ],
-                functions=function_definitions,
-                function_call="auto"
-            )
+            prompt = f"<|system|>\nYou are a helpful agent that answers questions correctly and concisely.\n<|user|>\n{question}\n<|assistant|>"
+            output = pipe(prompt, max_new_tokens=100, do_sample=False)
+            return output[0]["generated_text"].split("<|assistant|>")[-1].strip()
         except Exception as e:
-            print(f"LLM call failed: {e}")
             return f"LLM Error: {e}"
 
-        first_msg = first_resp.choices[0].message
-
-        if first_msg.function_call is not None:
-            fn_name = first_msg.function_call.name
-            fn_args_str = first_msg.function_call.arguments
-            try:
-                fn_args = json.loads(fn_args_str)
-            except json.JSONDecodeError:
-                return "Error: could not parse function_call arguments."
-
-            if fn_name == "python_exec":
-                code_to_run = fn_args.get("code", "")
-                python_output = python_exec(code_to_run)
-
-                try:
-                    second_resp = openai.chat.completions.create(
-                        model="gpt-4o-mini",
-                        messages=[
-                            {
-                                "role": "system",
-                                "content": "You are an assistant that can call python_exec when needed."
-                            },
-                            {"role": "user", "content": question},
-                            {
-                                "role": "assistant",
-                                "content": None,
-                                "function_call": {
-                                    "name": "python_exec",
-                                    "arguments": json.dumps({"code": code_to_run})
-                                }
-                            },
-                            {"role": "function", "name": "python_exec", "content": python_output}
-                        ]
-                    )
-                    return second_resp.choices[0].message.content.strip()
-                except Exception as e:
-                    return f"LLM Error: {e}"
-            else:
-                return f"Requested unknown function {fn_name}."
-        else:
-            return first_msg.content.strip()
-
 # ────────────────  API Helpers  ────────────────
 def get_all_questions(api_url: str) -> list[dict]:
     resp = requests.get(f"{api_url}/questions", timeout=15)
@@ -128,20 +49,16 @@ def submit_answers(api_url: str, username: str, code_link: str, answers: list[di
     return resp.json()
 
 # ────────────────  Gradio Evaluation Logic  ────────────────
-def run_and_submit_all(profile: Optional[gr.OAuthProfile] = None, *args):
-    if profile:
-        if isinstance(profile, dict):
-            username = profile.get("preferred_username", "").strip()
-        else:
-            username = str(profile).strip()
-    else:
+def run_and_submit_all(profile: gr.OAuthProfile | None):
+    if not profile:
         return "❌ Please log in to Hugging Face using the button above.", None
+    username = profile.username.strip()
 
     space_id = os.getenv("SPACE_ID", "")
     code_link = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else ""
 
     try:
-        agent = BasicAgent()
+        agent = ZephyrAgent()
     except Exception as e:
         return f"❌ Error initializing agent: {e}", None
 
@@ -152,6 +69,7 @@ def run_and_submit_all(profile: Optional[gr.OAuthProfile] = None, *args):
 
     answers_payload = []
     results_log = []
+
     for item in questions_data:
         task_id = item.get("task_id")
         question_text = item.get("question", "")
@@ -193,7 +111,7 @@ def run_and_submit_all(profile: Optional[gr.OAuthProfile] = None, *args):
 
 # ────────────────  Gradio UI  ────────────────
 with gr.Blocks() as demo:
-    gr.Markdown("# 🧠 Basic Agent Evaluation")
+    gr.Markdown("# 🧠 Zephyr-7B Agent Evaluation")
     gr.Markdown(
         """
         **Instructions:**
@@ -201,8 +119,6 @@ with gr.Blocks() as demo:
         1. Copy this Space and define your own agent logic.
         2. Log in with your Hugging Face account.
         3. Click “Run Evaluation & Submit All Answers” to test and submit.
-
-        - Format matters. Leaderboard uses exact match!
         """
     )
     gr.LoginButton()
@@ -210,28 +126,8 @@ with gr.Blocks() as demo:
     status_output = gr.Textbox(label="Status", lines=5, interactive=False)
     results_table = gr.DataFrame(label="Agent Answers", wrap=True)
 
-    run_button.click(
-        fn=run_and_submit_all,
-        inputs=[],
-        outputs=[status_output, results_table]
-    )
+    run_button.click(fn=run_and_submit_all, inputs=[], outputs=[status_output, results_table])
 
-# ────────────────  Launch App ────────────────
 if __name__ == "__main__":
-    space_host = os.getenv("SPACE_HOST")
-    space_id = os.getenv("SPACE_ID")
-
-    if space_host:
-        print(f"✅ SPACE_HOST found: {space_host}")
-        print(f"   Runtime URL: https://{space_host}.hf.space")
-    else:
-        print("ℹ️  SPACE_HOST not found (running locally)")
-
-    if space_id:
-        print(f"✅ SPACE_ID found: {space_id}")
-        print(f"   Repo URL: https://huggingface.co/spaces/{space_id}")
-    else:
-        print("ℹ️  SPACE_ID not found")
-
     print("Launching Gradio Interface...")
-    demo.launch(debug=True, server_name="0.0.0.0", server_port=7860)
+    demo.launch(debug=True, server_name="0.0.0.0", server_port=7860)