kaxil commented on code in PR #73024:
URL: https://github.com/apache/airflow/pull/73024#discussion_r3997906354
##########
providers/common/ai/docs/connections/pydantic_ai.rst:
##########
@@ -60,7 +60,7 @@ Host (optional)
- Ollama: ``http://localhost:11434/v1``
- vLLM: ``http://localhost:8000/v1``
- - Azure OpenAI:
``https://<resource>.openai.azure.com/openai/deployments/<deployment>``
+ - Azure OpenAI: ``https://<resource>.openai.azure.com``
Review Comment:
The generic `pydanticai` connection maps `host` to the provider's
`base_url`, and `AzureProvider` takes `azure_endpoint`, not `base_url` (checked
pydantic-ai 2.0.0, the floor in `pyproject.toml`, and 2.42.0). So with an
`azure:` model this host hits the `except TypeError` in `get_conn()`, which
warns and falls back to env-var auth, dropping the API key with it. With an
`openai:` model it does reach `OpenAIProvider(base_url=...)`, but Azure's
OpenAI-compatible surface lives at
`https://<resource>.openai.azure.com/openai/v1`, so the bare host 404s and the
old deployments path 400s on the missing `api-version`. `provider.yaml` also
lists Azure OpenAI only under `pydanticai_azure`, not under `pydanticai`. Could
this bullet point at the `pydanticai_azure` connection instead, or use the
`/openai/v1` form?
--
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]
