SEZ9 commented on PR #11548: URL: https://github.com/apache/seatunnel/pull/11548#issuecomment-5119889998
@yzeng1618 @davidzollo Thanks — all three findings addressed in head `640c2716f`: 1. **`--init` wiring** (davidzollo #1): the interactive setup now offers `bedrock-mantle` as choice 4 with shared AWS-credential guidance plus the `pip install ".[bedrock-mantle]"` note, and a dedicated model-selection branch (default `openai.gpt-5.6-terra`) that writes `OPENAI_MODEL*` and the matching config entry. The previous fall-through to the bedrock branch (writing `ANTHROPIC_MODEL*` that the provider never reads) is gone. 2. **Dependency floors** (davidzollo #2): `all` and `dev` now pin the same `openai>=2.45.0` floor as `bedrock-mantle` and both carry `aws-bedrock-token-generator`. Added a dependency-contract test that parses pyproject and asserts every extra bundling this provider has the required floor and token generator — so a future extra can't silently regress it. 3. **Setup-surface consistency** (davidzollo #3 / yzeng1618 docs note): the README `AI_PROVIDER` table and `env.example.sh` now list `bedrock-mantle` with its env contract. `setup.sh` installs `.[dev]`, which now carries the correct floor. On the docs/en + docs/zh point: the website AI CLI section (merged in #11551) already documents the provider in both locales under Quickstart → Provider Configuration; happy to expand that page with the region/token/`/openai/v1` detail as a small follow-up docs PR if you'd like it deeper than the current entry. Suite: 54 passed (17 provider tests + 3 routing tests + dependency contract). -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
