Release 0.15.1

[github-actions]

Release readiness review (v0.15.0 -> TARGET 93dc4ac)

This is a release readiness report done by $final-release-review skill.

Diff

v0.15.0...93dc4ac

Release call:

🟢 GREEN LIGHT TO SHIP Patch release scope is narrow, covered by focused tests, and no release-blocking regression, packaging break, or unversioned breaking change was identified.

Scope summary:

• 25 files changed (+0/-0); key areas touched: websocket Responses keepalive options, UnixLocal PTY signal handling, Dapr Redis integration fixture hardening, quickstart/release docs and translations, version bump to 0.15.1.
• Diff size: 1,111 insertions / 645 deletions, with most churn in generated/translated docs and focused runtime/test updates.

Risk assessment (ordered by impact):

No material risks identified.

1. Responses websocket keepalive options added to public configuration paths

   • Risk: 🟢 LOW. New optional parameters are appended and default to existing websocket behavior when omitted.
   • Evidence: OpenAIResponsesWebSocketOptions is added and exported; OpenAIProvider, MultiProvider, and responses_websocket_session pass it through only when supplied; OpenAIResponsesWSModel._open_websocket_connection conditionally adds ping_interval and ping_timeout.
   • Files: src/agents/models/openai_responses.py, src/agents/models/openai_provider.py, src/agents/models/multi_provider.py, src/agents/responses_websocket_session.py, src/agents/__init__.py
   • Action: Run uv run pytest -q tests/test_config.py tests/test_openai_responses.py; pass criteria: keepalive option pass-through tests and existing websocket reuse tests pass.

2. UnixLocal PTY child processes now restore SIGINT/SIGQUIT defaults

   • Risk: 🟢 LOW. This restores normal terminal behavior for PTY children when the parent process ignores signals, with targeted coverage for Ctrl-C and Ctrl-.
   • Evidence: _restore_pty_child_signal_defaults() is called in the PTY child preexec_fn, and tests/sandbox/test_unix_local.py asserts signal-driven exit codes for SIGINT and SIGQUIT while the parent ignores those signals.
   • Files: src/agents/sandbox/sandboxes/unix_local.py, tests/sandbox/test_unix_local.py
   • Action: Run uv run pytest -q tests/sandbox/test_unix_local.py; pass criteria: PTY signal regression tests pass on Unix CI.

3. Dapr Redis integration fixture changed component loading path and readiness checks

   • Risk: 🟢 LOW. The change affects integration test reliability rather than packaged runtime behavior.
   • Evidence: The fixture switches Dapr from -components-path to -resources-path, adjusts mounted file permissions, and waits for the statestore component through the metadata endpoint before proceeding.
   • Files: tests/extensions/memory/test_dapr_redis_integration.py
   • Action: In an environment with Docker and Dapr image access, run uv run pytest -q tests/extensions/memory/test_dapr_redis_integration.py; pass criteria: Redis state store component loads and integration tests pass.

4. Patch release includes documentation updates for prior 0.15.0 breaking-change notes

   • Risk: 🟢 LOW. Documentation-only additions improve migration visibility and do not alter runtime behavior.
   • Evidence: docs/release.md adds the 0.15.0 ModelRefusalError breaking-change note; docs/quickstart.md adds Windows activation and API key examples; translated docs are updated.
   • Files: docs/release.md, docs/quickstart.md, docs/ja/*, docs/ko/*, docs/zh/*
   • Action: Run make build-docs; pass criteria: MkDocs build completes without navigation or markdown errors.

Notes:

• Latest local release tag used: v0.15.0, per git tag -l 'v*' --sort=-v:refname | head -n1.
• Target commit used: 93dc4ac0b5c0d56d19ba756e6c7f4f46563afc41, per git rev-parse HEAD.
• Working tree status was clean.
• Local verification was not run as part of this review; this report assumes CI has run the repository’s normal release checks.