Add A Model Adapter
This guide describes how to add a provider without changing core orchestration.
1) Implement adapter
Create hexi.adapters.model_<provider>.py with adapter responsibilities only:
- auth/header handling,
- request formatting,
- response extraction.
Do not add orchestration logic in adapters.
2) Register provider mapping
Update provider resolution in hexi.cli._pick_model.
3) Support provider config block
Load [providers.<provider>] values through memory/config parsing.
4) Add tests
Required coverage:
- success response parsing,
- non-200/provider error handling,
- missing key failure path.
5) Validate end-to-end
Run:
PYTHONPATH=src pytest -q
hexi doctor --probe-model