T-124: Publish Exam AI to Hugging Face

This task involves setting up a manual workflow to publish the Exam AI model to the Hugging Face Hub.

Plan

1. Create a GitHub Actions workflow file .github/workflows/publish-to-hf.yml that runs only manually (workflow_dispatch).
2. The workflow will use the HF_TOKEN secret to push the model artifacts (adapter, merged weights, GGUF, WebLLM build) to the Hugging Face Hub.
3. The model card will be generated/updated to include:
   • License (Apache 2.0 or similar permissive).
   • Data sources (this list's content, original questions).
   • Evaluation results (from Phase 1).
   • Evidence-gate behavior (the model's tendency to cite sources or admit ignorance).
4. Verify the workflow configuration.

Notes

• The workflow must be manual only (no schedule).
• The model card must be comprehensive as per the requirements.
• I will document the process in ops/.notes.md.