Release 0.17.1

[github-actions]

Release readiness review (v0.17.0 -> TARGET 5ed3701)

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

Diff

v0.17.0...5ed3701

Release call:

🟢 GREEN LIGHT TO SHIP No concrete blocking regression, breaking public API change, data-loss issue, security regression, or release-critical packaging failure was identified.

Scope summary:

• 159 files changed (+7497/-1787); 8 added, 151 modified; 64 commits.
• Key areas touched: src/agents/models, src/agents/realtime, src/agents/run_internal, src/agents/sandbox, memory/session extensions, tracing, examples, tests, docs/translations, and version bump in pyproject.toml to 0.17.1.

Risk assessment (ordered by impact):

No material risks identified.

1. Chat Completions feature validation and filtering changes

   • Risk: 🟢 LOW. Default behavior remains compatibility-preserving by warning and ignoring unsupported Responses-only features, while strict failures are opt-in.
   • Evidence: OpenAIProvider and MultiProvider add keyword-only strict_feature_validation settings; OpenAIChatCompletionsModel now ignores prompt, previous_response_id, and conversation_id by default and raises only in strict mode. Tests cover non-streaming and streaming warnings/errors, multi-choice filtering, and custom tool-call rejection in strict mode.
   • Files: src/agents/models/openai_provider.py, src/agents/models/multi_provider.py, src/agents/models/openai_chatcompletions.py, src/agents/models/chatcmpl_converter.py, src/agents/models/chatcmpl_stream_handler.py, tests/models/test_openai_chatcompletions.py, tests/models/test_openai_chatcompletions_stream.py
   • Action: Run uv run pytest tests/models/test_openai_chatcompletions.py tests/models/test_openai_chatcompletions_stream.py; pass criteria: all focused Chat Completions tests pass.

2. Sandbox archive extraction validation was expanded

   • Risk: 🟢 LOW. Archive resource limits are opt-in by default, and the extraction path adds tests for preserving no-limit behavior, per-call/session limits, tar/zip byte limits, member limits, and invalid limit handling.
   • Evidence: SandboxArchiveLimits defaults are introduced but SandboxRunConfig.archive_limits defaults to None; tests assert the default preserves prior no-resource-limit behavior and validate configured limits.
   • Files: src/agents/run_config.py, src/agents/sandbox/session/archive_ops.py, src/agents/sandbox/session/archive_extraction.py, src/agents/sandbox/session/base_sandbox_session.py, tests/sandbox/test_extract.py, tests/sandbox/test_runtime.py
   • Action: Run uv run pytest tests/sandbox/test_extract.py tests/sandbox/test_runtime.py; pass criteria: archive extraction and runtime sandbox configuration tests pass.

3. Broad runtime/session/realtime fixes touch stateful paths

   • Risk: 🟢 LOW. The changes are mostly targeted bug fixes with matching regression tests across RunState serialization, session persistence, realtime cleanup, memory sessions, and tracing exporter resilience.
   • Evidence: New/updated tests cover guardrail payload serialization, OpenAI conversation item ID preservation, corrupt session item skipping, realtime iterator wakeup on close, realtime approval scoping, Redis metadata preservation, and tracing worker survival after exporter exceptions.
   • Files: src/agents/run_state.py, src/agents/run_internal/session_persistence.py, src/agents/realtime/session.py, src/agents/memory/sqlite_session.py, src/agents/extensions/memory/redis_session.py, src/agents/tracing/processors.py, tests/test_run_state.py, tests/memory/test_session_persistence_sanitize.py, tests/realtime/test_session.py, tests/test_trace_processor.py
   • Action: Run uv run pytest tests/test_run_state.py tests/memory/test_session_persistence_sanitize.py tests/realtime/test_session.py tests/test_trace_processor.py; pass criteria: all focused stateful runtime tests pass.

Notes:

• Base tag was selected from local tags only: v0.17.0.
• Target is current HEAD: 5ed37014769c50bb41e7f373a4dbc70a5f37de0c.
• Working tree status was clean.
• Local verification was not run as part of this review; this is not a release blocker under the $final-release-review gate policy.