potiuk opened a new pull request, #76312: URL: https://github.com/apache/airflow/pull/76312
Each pydantic-ai release that adds a provider module `infer_model` cannot route to — most recently the generic OpenAI-compatible provider in pydantic-ai-slim 2.63.0 — failed the `common.ai` `provider.yaml` drift tests until someone listed it in `UNREACHABLE_MODULES` by hand. That blocks the automated "Upgrade important CI environment" PR (#76278 needed a manual fixup for exactly this). Whether `infer_model` routes to a module can be asked of the installed pydantic-ai, trying both the module name and each provider's own `name` as the prefix. A module that only ever yields `Unknown model` is no vendor a connection can reach, so it is left out without a hand-kept list. A module that does route still has to be labelled and declared in `provider.yaml`, and one that starts routing in a later release now fails as unlabelled, which replaces the deleted guard test. The deliberate exclusions (`github`, `openai_codex`, `openai_chatgpt`) are unchanged. Tested the metadata tests on pydantic-ai-slim 2.62.0, 2.63.0 and the 2.36.0 floor, and checked that dropping a reachable vendor's label still fails them. --- ##### Was generative AI tooling used to co-author this PR? - [X] Yes (please specify the tool below) Generated-by: Claude Opus 5 following [the guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions) 🤖 Generated with [Claude Code](https://claude.com/claude-code) -- 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]
