m-bendik/agents-course-final-assignment

1import os2import gradio as gr3import requests4import pandas as pd5 6from smolagents import (7    CodeAgent,8    OpenAIServerModel,9)10from smolagents import CodeAgent, DuckDuckGoSearchTool, VisitWebpageTool11from dotenv import load_dotenv12import os13 14load_dotenv()15 16system_prompt = """17YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string.18"""19 20# model = OpenAIServerModel(model_id="GPT 4o mini",21#     api_base=os.environ["LITE_LLM_ENDPOINT"], api_key=os.environ["LITE_LLM_KEY"])22# model = OpenAIServerModel(model_id="o3")23model = OpenAIServerModel(model_id="gpt-4o")24agent = CodeAgent(tools=[DuckDuckGoSearchTool(), VisitWebpageTool()], model=model, add_base_tools=True, max_steps=30)25 26# (Keep Constants as is)27# --- Constants ---28DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"29 30# --- Basic Agent Definition ---31# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------32class BasicAgent:33    def __init__(self):34        print("BasicAgent initialized.")35    def __call__(self, question: str) -> str:36        print(f"Agent received question (first 50 chars): {question[:50]}...")37        answer = agent.run(question + "\n\n" + system_prompt)38        print(f"Agent returning fixed answer: {answer}")39        return answer40 41def run_and_submit_all( profile: gr.OAuthProfile | None):42    """43    Fetches all questions, runs the BasicAgent on them, submits all answers,44    and displays the results.45    """46    # --- Determine HF Space Runtime URL and Repo URL ---47    space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code48 49    if profile:50        username= f"{profile.username}"51        print(f"User logged in: {username}")52    else:53        print("User not logged in.")54        return "Please Login to Hugging Face with the button.", None55 56    api_url = DEFAULT_API_URL57    questions_url = f"{api_url}/questions"58    submit_url = f"{api_url}/submit"59 60    # 1. Instantiate Agent ( modify this part to create your agent)61    try:62        agent = BasicAgent()63    except Exception as e:64        print(f"Error instantiating agent: {e}")65        return f"Error initializing agent: {e}", None66    # In the case of an app running as a hugging Face space, this link points toward your codebase ( usefull for others so please keep it public)67    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"68    print(agent_code)69 70    # 2. Fetch Questions71    print(f"Fetching questions from: {questions_url}")72    try:73        response = requests.get(questions_url, timeout=15)74        response.raise_for_status()75        questions_data = response.json()76        if not questions_data:77             print("Fetched questions list is empty.")78             return "Fetched questions list is empty or invalid format.", None79        print(f"Fetched {len(questions_data)} questions.")80    except requests.exceptions.RequestException as e:81        print(f"Error fetching questions: {e}")82        return f"Error fetching questions: {e}", None83    except requests.exceptions.JSONDecodeError as e:84         print(f"Error decoding JSON response from questions endpoint: {e}")85         print(f"Response text: {response.text[:500]}")86         return f"Error decoding server response for questions: {e}", None87    except Exception as e:88        print(f"An unexpected error occurred fetching questions: {e}")89        return f"An unexpected error occurred fetching questions: {e}", None90 91    # 3. Run your Agent92    results_log = []93    answers_payload = []94    print(f"Running agent on {len(questions_data)} questions...")95    for item in questions_data:96        task_id = item.get("task_id")97        question_text = item.get("question")98        if not task_id or question_text is None:99            print(f"Skipping item with missing task_id or question: {item}")100            continue101        try:102            submitted_answer = agent(question_text)103            answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})104            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})105        except Exception as e:106             print(f"Error running agent on task {task_id}: {e}")107             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})108 109    if not answers_payload:110        print("Agent did not produce any answers to submit.")111        return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)112 113    # 4. Prepare Submission 114    submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}115    status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."116    print(status_update)117 118    # 5. Submit119    print(f"Submitting {len(answers_payload)} answers to: {submit_url}")120    try:121        response = requests.post(submit_url, json=submission_data, timeout=60)122        response.raise_for_status()123        result_data = response.json()124        final_status = (125            f"Submission Successful!\n"126            f"User: {result_data.get('username')}\n"127            f"Overall Score: {result_data.get('score', 'N/A')}% "128            f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"129            f"Message: {result_data.get('message', 'No message received.')}"130        )131        print("Submission successful.")132        results_df = pd.DataFrame(results_log)133        return final_status, results_df134    except requests.exceptions.HTTPError as e:135        error_detail = f"Server responded with status {e.response.status_code}."136        try:137            error_json = e.response.json()138            error_detail += f" Detail: {error_json.get('detail', e.response.text)}"139        except requests.exceptions.JSONDecodeError:140            error_detail += f" Response: {e.response.text[:500]}"141        status_message = f"Submission Failed: {error_detail}"142        print(status_message)143        results_df = pd.DataFrame(results_log)144        return status_message, results_df145    except requests.exceptions.Timeout:146        status_message = "Submission Failed: The request timed out."147        print(status_message)148        results_df = pd.DataFrame(results_log)149        return status_message, results_df150    except requests.exceptions.RequestException as e:151        status_message = f"Submission Failed: Network error - {e}"152        print(status_message)153        results_df = pd.DataFrame(results_log)154        return status_message, results_df155    except Exception as e:156        status_message = f"An unexpected error occurred during submission: {e}"157        print(status_message)158        results_df = pd.DataFrame(results_log)159        return status_message, results_df160 161 162# --- Build Gradio Interface using Blocks ---163with gr.Blocks() as demo:164    gr.Markdown("# Basic Agent Evaluation Runner")165    gr.Markdown(166        """167        **Instructions:**168        1.  Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...169        2.  Log in to your Hugging Face account using the button below. This uses your HF username for submission.170        3.  Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.171        ---172        **Disclaimers:**173        Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).174        This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.175        """176    )177 178    gr.LoginButton()179 180    run_button = gr.Button("Run Evaluation & Submit All Answers")181 182    status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)183    # Removed max_rows=10 from DataFrame constructor184    results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)185 186    run_button.click(187        fn=run_and_submit_all,188        outputs=[status_output, results_table]189    )190 191if __name__ == "__main__":192    print("\n" + "-"*30 + " App Starting " + "-"*30)193    # Check for SPACE_HOST and SPACE_ID at startup for information194    space_host_startup = os.getenv("SPACE_HOST")195    space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup196 197    if space_host_startup:198        print(f"✅ SPACE_HOST found: {space_host_startup}")199        print(f"   Runtime URL should be: https://{space_host_startup}.hf.space")200    else:201        print("ℹ️  SPACE_HOST environment variable not found (running locally?).")202 203    if space_id_startup: # Print repo URLs if SPACE_ID is found204        print(f"✅ SPACE_ID found: {space_id_startup}")205        print(f"   Repo URL: https://huggingface.co/spaces/{space_id_startup}")206        print(f"   Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")207    else:208        print("ℹ️  SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")209 210    print("-"*(60 + len(" App Starting ")) + "\n")211 212    print("Launching Gradio Interface for Basic Agent Evaluation...")213    demo.launch(debug=True, share=False)214