Spaces:
Runtime error
Runtime error
| #!/usr/bin/env python3 | |
| """ | |
| 🚀 Enhanced GAIA Agent Interface - Full API Integration | |
| Complete Gradio interface for GAIA benchmark with API connectivity and scoring | |
| """ | |
| import os | |
| import gradio as gr | |
| import json | |
| from datetime import datetime | |
| from gaia_agent import ModularGAIAAgent | |
| agent = ModularGAIAAgent() | |
| def run_api_questions(): | |
| results = agent.run(from_api=True) | |
| answers = "" | |
| for r in results: | |
| answers += f"Task ID: {r['task_id']}\nAnswer: {r['answer']}\nReasoning Trace: {' | '.join(r['reasoning_trace'])}\n\n" | |
| return answers | |
| def run_manual_question(question): | |
| qobj = {"task_id": "manual", "question": question, "file_name": ""} | |
| answer, trace = agent.answer_question(qobj) | |
| return answer, "\n".join(trace) | |
| def show_help(): | |
| return ( | |
| "# Agent Capabilities\n" | |
| "- Multi-modal QA (text, audio, image, code, table, YouTube/video)\n" | |
| "- File download and analysis from API\n" | |
| "- Advanced video QA: object detection, captioning, ASR\n" | |
| "- Secure code execution\n" | |
| "- Robust error handling and logging\n" | |
| "- GAIA-compliant output\n" | |
| "\nSee README.md for full details." | |
| ) | |
| def submit_answers(username, agent_code_url): | |
| # Placeholder for submission logic | |
| return f"Submission for {username} with code {agent_code_url} (not implemented in demo)" | |
| def show_leaderboard(): | |
| # Placeholder for leaderboard logic | |
| return "Leaderboard feature coming soon." | |
| demo = gr.Blocks(title="GAIA Benchmark Agent", theme=gr.themes.Soft()) | |
| with demo: | |
| gr.Markdown(""" | |
| # 🤖 GAIA Benchmark Agent | |
| Multi-modal, multi-step reasoning agent for the Hugging Face GAIA benchmark. | |
| """) | |
| with gr.Tabs(): | |
| with gr.TabItem("API Q&A"): | |
| api_btn = gr.Button("Run on API Questions", variant="primary") | |
| api_output = gr.Textbox(label="Answers and Reasoning Trace", lines=20) | |
| api_btn.click(run_api_questions, outputs=api_output) | |
| with gr.TabItem("Manual Input"): | |
| manual_q = gr.Textbox(label="Enter your question", lines=3) | |
| manual_btn = gr.Button("Answer", variant="primary") | |
| manual_a = gr.Textbox(label="Answer") | |
| manual_trace = gr.Textbox(label="Reasoning Trace", lines=5) | |
| manual_btn.click(run_manual_question, inputs=manual_q, outputs=[manual_a, manual_trace]) | |
| with gr.TabItem("Submission/Leaderboard"): | |
| username = gr.Textbox(label="Hugging Face Username") | |
| code_url = gr.Textbox(label="Agent Code URL") | |
| submit_btn = gr.Button("Submit Answers", variant="primary") | |
| submit_out = gr.Textbox(label="Submission Result") | |
| submit_btn.click(submit_answers, inputs=[username, code_url], outputs=submit_out) | |
| leaderboard_btn = gr.Button("Show Leaderboard") | |
| leaderboard_out = gr.Textbox(label="Leaderboard") | |
| leaderboard_btn.click(show_leaderboard, outputs=leaderboard_out) | |
| with gr.TabItem("Agent Help"): | |
| help_md = gr.Markdown(show_help()) | |
| if __name__ == "__main__": | |
| demo.launch() | |