import os import inspect import requests import pandas as pd import gradio as gr from smolagents import CodeAgent, LiteLLMModel, DuckDuckGoSearchTool, VisitWebpageTool, tool # --- Constants --- DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" # --------------------------------------------------------------------------- # Custom tools for file-based question types (audio, images, python files) # --------------------------------------------------------------------------- @tool def transcribe_audio(file_path: str) -> str: """ Transcribes an audio file (mp3/wav) to text using OpenAI Whisper. Args: file_path: Local path to the audio file to transcribe. Returns: The transcribed text. """ from openai import OpenAI client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY")) with open(file_path, "rb") as f: transcript = client.audio.transcriptions.create( model="whisper-1", file=f ) return transcript.text @tool def analyze_image(file_path: str, question: str) -> str: """ Analyzes an image (e.g. a chess position) using a vision-capable LLM and answers a question about it. Args: file_path: Local path to the image file. question: The question to answer about the image. Returns: The model's answer about the image. """ import base64 from openai import OpenAI client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY")) with open(file_path, "rb") as f: b64_image = base64.b64encode(f.read()).decode("utf-8") response = client.chat.completions.create( model="gpt-4o", messages=[{ "role": "user", "content": [ {"type": "text", "text": question}, {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{b64_image}"}} ] }] ) return response.choices[0].message.content @tool def run_python_file(file_path: str) -> str: """ Reads and returns the contents of a Python (.py) file so the agent can analyze or trace through the code to determine its output. Args: file_path: Local path to the python file. Returns: The raw source code as text. """ with open(file_path, "r") as f: return f.read() @tool def read_excel_file(file_path: str) -> str: """ Reads an Excel (.xlsx) file and returns its contents as a string table. Args: file_path: Local path to the Excel file. Returns: A string representation of the spreadsheet data. """ df = pd.read_excel(file_path) return df.to_string() # --------------------------------------------------------------------------- # The Agent # --------------------------------------------------------------------------- class BasicAgent: def __init__(self): print("BasicAgent initialized.") # LiteLLM lets you swap providers by just changing model_id, e.g.: # "gpt-4o-mini", "claude-3-5-sonnet-20241022", "huggingface/Qwen/Qwen2.5-72B-Instruct" self.model = LiteLLMModel( model_id="gpt-4o-mini", api_key=os.environ.get("OPENAI_API_KEY"), ) self.agent = CodeAgent( model=self.model, tools=[ DuckDuckGoSearchTool(), VisitWebpageTool(), transcribe_audio, analyze_image, run_python_file, read_excel_file, ], max_steps=8, ) def __call__(self, question: str, file_path: str = None) -> str: print(f"Agent received question (first 80 chars): {question[:80]}...") prompt = question if file_path: prompt += f"\n\nA file has been downloaded for this question at local path: {file_path}. Use the appropriate tool to read it before answering." prompt += "\n\nIMPORTANT: Respond with ONLY the final answer. No explanation, no 'FINAL ANSWER:' prefix, just the answer itself, formatted exactly as requested in the question." try: answer = self.agent.run(prompt) except Exception as e: print(f"Agent error: {e}") answer = "ERROR" answer = str(answer).strip() print(f"Agent returning answer: {answer}") return answer # --------------------------------------------------------------------------- # Evaluation + submission logic # --------------------------------------------------------------------------- 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: return "Please log in to Hugging Face first.", None api_url = DEFAULT_API_URL questions_url = f"{api_url}/questions" files_url = f"{api_url}/files" submit_url = f"{api_url}/submit" agent = BasicAgent() agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" # 1. Fetch questions try: response = requests.get(questions_url, timeout=15) response.raise_for_status() questions_data = response.json() 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") question_text = item.get("question") file_name = item.get("file_name", "") if not task_id or question_text is None: continue file_path = None if file_name: try: file_resp = requests.get(f"{files_url}/{task_id}", timeout=30) file_resp.raise_for_status() file_path = f"/tmp/{file_name}" with open(file_path, "wb") as f: f.write(file_resp.content) except Exception as e: print(f"Could not download file for {task_id}: {e}") try: submitted_answer = agent(question_text, file_path) except Exception as e: submitted_answer = f"AGENT ERROR: {e}" 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}) if not answers_payload: return "No answers were generated.", pd.DataFrame(results_log) # 2. Submit submission_data = { "username": username.strip(), "agent_code": agent_code, "answers": answers_payload } 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', '')}" ) return final_status, pd.DataFrame(results_log) except Exception as e: return f"Submission failed: {e}", pd.DataFrame(results_log) # --------------------------------------------------------------------------- # Gradio UI # --------------------------------------------------------------------------- with gr.Blocks() as demo: gr.Markdown("# Basic Agent Evaluation Runner") gr.Markdown( """ **Instructions:** 1. This Space defines your agent's logic, tools, and required packages. 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 the score. """ ) 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__": demo.launch(debug=True, share=False)