import os import gradio as gr import requests import pandas as pd from smolagents import InferenceClientModel, ToolCallingAgent from audio_transcriber import AudioTranscriptionTool from image_analyzer import ImageAnalysisTool from wikipedia_searcher import WikipediaSearcher DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" # Zephyr-compatible system prompt to prepend manually SYSTEM_PROMPT = ( "You are an agent solving the GAIA benchmark and must provide exact answers.\n" "Rules:\n" "1. Return only the exact requested answer: no explanation.\n" "2. For yes/no, return 'Yes' or 'No'.\n" "3. For dates, use the exact requested format.\n" "4. For numbers, use only the number.\n" "5. For names, use the exact name from sources.\n" "6. If the question has a file, download it using the task ID.\n" "Examples:\n" "- '42'\n" "- 'Arturo Nunez'\n" "- 'Yes'\n" "- 'October 5, 2001'\n" "- 'Buenos Aires'\n" "Never say 'the answer is...'. Only return the answer.\n" ) class GaiaAgent: def __init__(self): print("Gaia Agent Initialized") self.model = InferenceClientModel( model_id="HuggingFaceH4/zephyr-7b-beta", token=os.getenv("HF_API_TOKEN", "").strip() ) self.tools = [ AudioTranscriptionTool(), ImageAnalysisTool(), WikipediaSearcher() ] self.agent = ToolCallingAgent( tools=self.tools, model=self.model ) def __call__(self, question: str) -> str: print(f"Agent received question (first 50 chars): {question[:50]}...") full_prompt = f"{SYSTEM_PROMPT}\nQUESTION:\n{question}" try: result = self.agent.run(full_prompt) print(f"Raw result from agent: {result}") if isinstance(result, dict) and "answer" in result: return str(result["answer"]).strip() elif isinstance(result, str): return result.strip() elif isinstance(result, list): for item in reversed(result): if isinstance(item, dict) and item.get("role") == "assistant" and "content" in item: return item["content"].strip() return "ERROR: Unexpected list format" else: return "ERROR: Unexpected result type" except Exception as e: print(f"Exception during agent run: {e}") return f"AGENT ERROR: {e}" def run_and_submit_all(profile: gr.OAuthProfile | None): space_id = os.getenv("SPACE_ID") if profile: username = profile.username print(f"User logged in: {username}") else: print("User not logged in.") 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" try: agent = GaiaAgent() except Exception as e: print(f"Error initializing agent: {e}") return f"Error initializing agent: {e}", None agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" print(f"Agent code URL: {agent_code}") try: response = requests.get(questions_url, timeout=15) 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 results_log = [] answers_payload = [] for item in questions_data: task_id = item.get("task_id") if not task_id: continue try: submitted_answer = agent(item.get("question", "")) answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer}) results_log.append({ "Task ID": task_id, "Question": item.get("question", ""), "Submitted Answer": submitted_answer }) except Exception as e: error_msg = f"AGENT ERROR: {e}" results_log.append({ "Task ID": task_id, "Question": item.get("question", ""), "Submitted Answer": error_msg }) if not answers_payload: return "Agent did not produce any answers to submit.", pd.DataFrame(results_log) submission_data = { "username": username.strip(), "agent_code": agent_code, "answers": answers_payload } print(f"Submitting {len(answers_payload)} answers to: {submit_url}") try: response = requests.post(submit_url, json=submission_data, timeout=60) response.raise_for_status() result_data = response.json() final_status = ( f"Submission Successful!\n" f"User: {result_data.get('username')}\n" f"Overall Score: {result_data.get('score', 'N/A')}% " 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 except requests.exceptions.HTTPError as e: try: detail = e.response.json().get("detail", e.response.text) except Exception: detail = e.response.text[:500] return f"Submission Failed: {detail}", pd.DataFrame(results_log) except requests.exceptions.Timeout: return "Submission Failed: The request timed out.", pd.DataFrame(results_log) except Exception as e: return f"An unexpected error occurred during submission: {e}", pd.DataFrame(results_log) # Gradio UI with gr.Blocks() as demo: gr.Markdown("# Basic Agent Evaluation Runner") gr.Markdown(""" **Instructions:** 1. Clone this space and define your agent and tools. 2. Log in to your Hugging Face account using the button below. 3. Click 'Run Evaluation & Submit All Answers' to test your agent and submit results. """) 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]) if __name__ == "__main__": print("\n" + "-"*30 + " App Starting " + "-"*30) 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 should be: https://{space_host}.hf.space") else: print("ℹ️ SPACE_HOST not found.") 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("-"*(60 + len(" App Starting ")) + "\n") demo.launch(debug=True, share=False)