[azure-ai-projects] Emit SDK and add samples and tests

[Glenn Harper (glharper)]

Summary

• emit azure-ai-projects from Azure/azure-rest-api-specs commit 592d36dbd539e1365d0e8c0e993daf8ffeda29ef
• apply package post-emitter fixes, preview feature-header wiring, and regenerate APIView stubs
• add synchronous and asynchronous WebIQ and shell-toolbox samples
• add complete recorded service tests that remain skipped until recordings are added
• update the 2.6.0 changelog

Skills used

• azure-ai-projects-emit-from-typespec
• azure-ai-projects-author-samples
• azure-ai-projects-author-tests
• azure-ai-projects-update-changelog

This branch was based on the authoring-skills branch, so it also contains the skill definitions submitted separately in #48772.

Validation

• 190 preview feature-header tests passed
• 4 new recorded service tests collect and skip as intended
• pylint passed
• source and newly added sample mypy checks passed

The full sample mypy run still reports pre-existing OpenAI union-typing errors in three multi-agent samples; none involve the new files.

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).
10 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[github-actions]

[Pilot] PR Pipeline Failure Analysis

What failed

Pipeline: python - pullrequest (Build ID 6750191)

All three test platforms failed with the same root cause — playback failure due to a <Foundry-Features> header mismatch between recorded sessions and the current SDK:

Platform	Failed tests (approx.)
macos311 (whl + sdist)	~136
Ubuntu2404_313 (whl + sdist)	~136
windows2022_312 (whl + sdist)	~136

Three failure groups, one root cause:

1. Telemetry tests (test_ai_agents_instrumentor, test_responses_instrumentor, test_responses_instrumentor_workflow, and their _async variants) — 16 tests × platforms
2. Agent CRUD tests (test_agent_disable_enable, test_prompt_agent_endpoint_responses, test_prompt_agent_endpoint_responses_async) — 3 tests × platforms
3. Sample tests (test_agent_tools_samples, test_agents_samples, test_hosted_agents_samples, and their _async variants) — all sample tests × platforms

Relevant pipeline output

azure.core.exceptions.ResourceNotFoundError: Playback failure -- for help resolving,
see https://aka.ms/azsdk/python/test-proxy/troubleshoot. Error details:
Unable to find a record for the request POST
(sanitized.services.ai.azure.com/redacted)
test-workflow-agent/versions?api-version=v1
Header differences:
    <Foundry-Features> values differ,
    request <WorkflowAgents=V1Preview,ExternalAgents=V1Preview,DraftAgents=V1Preview,
              AgentsOptimization=V2Preview,ModelRouterControls=V1Preview>,
    record  <WorkflowAgents=V1Preview,ExternalAgents=V1Preview,DraftAgents=V1Preview,
              AgentsOptimization=V2Preview>
Body differences:
Remaining Entries:
0: (sanitized/redacted)
1: (sanitized/redacted)

Root cause

The new TypeSpec emit added ModelRouterControls=V1Preview to the <Foundry-Features> request header. Every existing test recording was captured without this feature flag, so test-proxy cannot match the incoming requests to the stored cassettes and returns 404. This causes all 136+ playback-mode tests to fail identically across all three platforms and both install formats (whl, sdist).

Recommended next steps

• Re-record all affected test sessions with live Azure credentials so recordings include the new ModelRouterControls=V1Preview feature flag. Run tests in record mode:

  cd sdk/ai/azure-ai-projects
  AZURE_TEST_RUN_LIVE=true pytest tests/agents/telemetry/ tests/agents/test_agents_crud.py tests/samples/ --record-mode=record

• Commit the updated recording files in sdk/ai/azure-ai-projects/tests/recordings/.
• Alternatively, if ModelRouterControls=V1Preview should be excluded from header matching in the test proxy, add a sanitizer/matcher configuration to strip or ignore this header during playback.
• See https://aka.ms/ci-fix

Copilot detected the failing pipeline and generated the analysis above. To have it attempt a fix automatically, reply with @copilot please fix the failing pipeline on this PR.

Generated by Pipeline Analysis Next Steps · 70.3 AIC · ⌖ 9.07 AIC · ⊞ 6K · ◷