potiuk opened a new pull request, #71520: URL: https://github.com/apache/airflow/pull/71520
Reverts the workaround from #71273, which dropped the `evaluation` extra of `google-cloud-aiplatform` and inlined its contents, because that extra capped litellm below the versions we need on Python 3.14. `google-cloud-aiplatform` 1.164.0 lifts the cap — its `evaluation` extra now asks for `litellm>=1.83.7,<1.86.0` under 3.14 and `litellm>=1.93.0,<1.97.0` from 3.14 on, which is what the inlined copy asked for. Verified on PyPI: 1.164.0 was published 2026-08-12 and its `requires_dist` carries those two markers. The floor moves to 1.164.0 because every earlier release still ships the capped extra. The `uv.lock` resolution is unchanged for litellm (1.85.7 under 3.14, 1.95.0 from 3.14 on). scikit-learn on Python 3.10 goes back to 1.5.2, which is what the upstream extra pins there (`scikit-learn<1.6.0; python_version<='3.10'`) — the `>=1.6.0` floor existed only in the inlined copy. 1.164.0 was published inside the rolling `exclude-newer = "4 days"` window, so it needs a temporary per-package cutoff in the root `pyproject.toml`, following the same pattern as the `pydantic-ai-skills` entry next to it. That entry can go once the window advances past 2026-08-12. Fixed upstream in: https://github.com/googleapis/python-aiplatform/issues/7057 --- ##### Was generative AI tooling used to co-author this PR? - [X] Yes — Claude Code (Opus 5) Generated-by: Claude Code (Opus 5) following [the guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions) -- 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]
