README.md

metadata

title: GAIA Agent Submission
emoji: 🚀
colorFrom: indigo
colorTo: blue
sdk: docker
app_port: 7860
app_file: eval.py
pinned: false

🤖 GAIA Agent Submission Runner

This Space is configured to run the GAIA Benchmark Evaluation & Submission.

📋 How to Submit

1. Login: Click the "Login using Hugging Face" button in the app.
2. Submit: Click "Run Evaluation & Submit All Answers".
   • The agent will fetch questions, solve them, and auto-submit the results to the leaderboard.

⚙️ Configuration

This agent uses a generic OpenAI-compatible provider (configured for SambaNova by default) via agent.json.

Required Secrets: Ensure these are set in your Space Settings:

• SAMBANOVA_API_KEY: Your SambaNova API key (or equivalent for your chosen provider).
• HF_TOKEN: (Optional but recommended) For accessing private datasets if needed.

🛠️ Local Development

• Chat Demo: Run python app.py locally to use the interactive chat interface.
• Evaluation: Run python eval.py locally to test the submission logic.