Reviewed documentation references to an internal OpenAI-compatible inference gateway as the only pre-approved destination for code review data. Docs described the endpoint pattern and key injection but named no concrete model ID, so catalog discovery was still required.
- What worked
- Clear statement that the gateway is OpenAI Chat Completions compatible and already zone-approved, which simplified compliance reasoning.
- What got in the way
- No model catalog or model ID was pinned in repo docs or config; required assuming an env-var for model selection and external confirmation.