kaxil opened a new pull request, #62824: URL: https://github.com/apache/airflow/pull/62824
## Summary Adds `providers/common/ai/AGENTS.md` to guide AI coding tools and contributors toward correct design decisions when working on the common AI provider. ## Why PR #62816 added ~280 lines of Azure OpenAI builder/factory code that duplicated what pydantic-ai's `AzureProvider` already handles natively. An AGENTS.md in the provider directory would have steered the contributor (and AI coding tools like Cursor, Claude Code, Copilot) toward the simpler approach from the start. ## What the file covers - **Design principles**: delegate to pydantic-ai, keep the hook thin, no premature abstractions - **Adding new LLM providers**: check pydantic-ai first, only add a `get_conn()` branch if needed - **Security**: no dynamic imports from connection extras, SQL validation stays on by default - **Common pitfalls**: don't construct raw SDK clients, don't add provider-specific connection types, use `common.compat` for SDK imports - **Key file paths** for quick navigation Follows the style of `airflow-core/src/airflow/api_fastapi/execution_api/AGENTS.md` — domain-specific and concise, not generic setup. The root `providers/AGENTS.md` already covers general provider conventions. -- 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]
