GAIA Agent Evaluation Runner (Unit 4)

Workflow

1. Log in with Hugging Face (OAuth).
2. Test on Random Question — sanity-check answer formatting before submitting.
3. Run Evaluation & Submit All Answers — runs all 20 questions and posts to the course scorer.

Requires Space secret GROQ_API_KEY. For file attachments, also set HF_TOKEN (with GAIA dataset access). On free ZeroGPU, do not set ZEROGPU_TASK_DURATION (use app default).