GAIA Level-1 Agent – Evaluation Runner

1. Log in with your Hugging Face account.
2. Run agent answers all 20 questions (takes several minutes). Answers are cached per question, so re-running only fills in what is missing.
3. Submit answers sends the cached answers for scoring and updates the leaderboard.

Model is chosen with env vars: MODEL_ID (LiteLLM id + provider API key) or HF_MODEL_ID + HF_TOKEN. See agent.py.