regression URL parsing in PR #231

[bmerkle]

fixes bugs described in #238

• regression URL parsing in PR Fix parse_azure_endpoint passing query string to AsyncAzureOpenAI #231
• make format on main

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Fixes the Azure OpenAI endpoint parsing regression reported in #238 (introduced in PR #231) and applies formatting/cleanup across tooling, library code, and tests.

Changes:

• Update Azure endpoint parsing to strip query strings and deployment paths (tests updated accordingly).
• Add an Ocp-Apim-Subscription-Key default header to Azure OpenAI client construction paths.
• Apply formatting/import cleanup across tools, library modules, and tests.

Reviewed changes

Copilot reviewed 12 out of 13 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tools/release.py	Reformat print(..., file=sys.stderr) calls for consistency.
tools/query.py	Reformat/refactor typeagent.knowpro imports.
src/typeagent/aitools/utils.py	Strip deployment path from Azure endpoint; add APIM subscription header; adjust agent Azure wiring.
src/typeagent/aitools/model_adapters.py	Add APIM subscription header to Azure provider client.
src/typeagent/knowpro/conversation_base.py	Reformat/refactor imports (including knowledge_schema as kplib).
src/typeagent/knowpro/knowledge.py	Reformat imports.
src/typeagent/storage/memory/semrefindex.py	Reformat imports.
tests/test_utils.py	Update expectations/docs for endpoint parsing (base host only).
tests/test_online.py	Switch online test to the create_chat_model (pydantic-ai) execution path.
tests/test_semrefindex.py	Reformat imports.
tests/test_knowledge.py	Reformat imports.
tests/benchmarks/test_benchmark_vectorbase.py	Reformat imports and a long function signature.

Comments suppressed due to low confidence (1)

src/typeagent/aitools/utils.py:1

• Setting Ocp-Apim-Subscription-Key unconditionally to the Azure OpenAI API key is risky and potentially incorrect: this header is typically for Azure API Management (APIM) subscription keys, not for direct Azure OpenAI. It also duplicates sending the secret in an additional header. Consider making this header opt-in (e.g., only set when a dedicated env var like AZURE_APIM_SUBSCRIPTION_KEY is present) and keep Azure OpenAI auth confined to the standard api_key mechanism.

# Copyright (c) Microsoft Corporation.

[KRRT7]

Two edge cases to address before merging:

1. Bare /openai suffix not stripped — If an endpoint looks like https://foo.openai.azure.com/openai?api-version=..., the current fix leaves /openai intact. AsyncAzureOpenAI.__init__ unconditionally appends /openai to the base URL (see openai/lib/azure.py:226-228), producing https://foo.openai.azure.com/openai/openai/... → 404. Trivial fix — add a re.sub(r"/openai$", "", clean_endpoint) after the deployment path stripping.

2. AZURE_APIM_SUBSCRIPTION_KEY header missing from the identity branch — The APIM header is only added in the else (raw key) branch of _make_azure_provider. If someone uses managed identity behind Azure API Management, the Ocp-Apim-Subscription-Key header is silently omitted. Lift the apim_key lookup and default_headers construction above the if/else so both branches get it.

[KRRT7]

This handles /openai/deployments/{name} but misses a bare /openai suffix. AsyncAzureOpenAI always appends /openai to the base URL (see openai/lib/azure.py:226-228), so an endpoint like https://foo.openai.azure.com/openai?api-version=... would produce /openai/openai/....

Suggested addition after this line:

# Also strip a bare /openai suffix — the SDK always appends /openai.
clean_endpoint = re.sub(r"/openai$", "", clean_endpoint)

[bmerkle]

i think we should use:

   # Strip query string and /openai... path — AsyncAzureOpenAI expects a
    # clean base URL and builds the deployment path internally.
    clean_endpoint = azure_endpoint.split("?", 1)[0]
    clean_endpoint = re.sub(r"/openai(/deployments/.*)?$", "", clean_endpoint)

[gvanrossum]

@KRRT7 Can you review and either approve or add feedback for @bmerkle? Once you two agree, and no sooner, can I review/skim and merge.