Skip to content

regression URL parsing in PR #231 #238

@bmerkle

Description

@bmerkle

Hej, @KRRT7 @gvanrossum

Seems we have a regression with the PR #231
#231
02ca2d2

My .env file contains AZURE_OPENAI_ENDPOINT and AZURE_OPENAI_ENDPOINT_EMBEDDING which includes the api version after ?
Note that this is default, how you get it from the foundry UI when creating new deployments, to this format should be the default in .env files and it should also work out of the box. (as most users will use this format)

make test now fails with my .env files containing ? and api version

we have two problems:

  1. AsyncAzureOpenAI then appends its own /openai/deployments/gpt-4o/chat/completions, resulting in a doubled path → 404.
  2. The test_online.py test uses create_async_openai_client (a different code path in utils.py), whereas the failing test goes through create_chat_model → _make_azure_provider → AzureProvider(openai_client=AsyncAzureOpenAI(...)).

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions