Release 0.13.0

[github-actions]

Release readiness review (v0.12.5 -> TARGET ac53702)

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

Diff

v0.12.5...ac53702

Release call:

🟢 GREEN LIGHT TO SHIP Version bump to 0.13.0 matches the user-visible changes; no concrete release-blocking regressions were identified in the reviewed diff.

Scope summary:

• 73 files changed (+3936/-584); key areas touched: MCP server/runtime APIs, chat-completions reasoning replay, realtime model defaults, session/memory race fixes, tool execution race handling, extensive targeted test additions, and docs updates.

Risk assessment (ordered by impact):

1. Default realtime model changed

• Risk: 🟡 MODERATE. Users relying on implicit defaults may see behavior/performance differences because websocket default moved from gpt-realtime to gpt-realtime-1.5.
• Evidence: src/agents/realtime/openai_realtime.py default model constant and constructor/session fallback changed; corresponding assertions added in tests/realtime/test_openai_realtime.py; docs updated in docs/realtime/guide.md and docs/realtime/quickstart.md.
• Files: src/agents/realtime/openai_realtime.py, tests/realtime/test_openai_realtime.py, docs/realtime/guide.md, docs/realtime/quickstart.md
• Action: Run uv run pytest -q tests/realtime/test_openai_realtime.py -k "gpt_realtime_1_5 or session_config_defaults_audio_formats_when_not_call" and confirm release notes explicitly call out the default-model change.

2. MCP public surface expansion (resources + session_id)

• Risk: 🟢 LOW. Changes are additive and covered, but they broaden runtime behavior and integration paths.
• Evidence: New list_resources, list_resource_templates, read_resource methods and MCPServerStreamableHttp.session_id property added in src/agents/mcp/server.py; dedicated tests added for resources and session ID.
• Files: src/agents/mcp/server.py, tests/mcp/test_mcp_resources.py, tests/mcp/test_streamable_http_session_id.py
• Action: Run uv run pytest -q tests/mcp/test_mcp_resources.py tests/mcp/test_streamable_http_session_id.py tests/mcp/test_mcp_server_manager.py; pass criteria is all tests green with no new skips/failures.

3. Stateful runtime race-condition fixes (memory/tool execution/compaction)

• Risk: 🟢 LOW. Changes target known race/error paths and are backed by regression tests, with no direct breaking API signal.
• Evidence: Race-safe first writes and lock handling in src/agents/extensions/memory/sqlalchemy_session.py; eager-task cancellation/state fix in src/agents/run_internal/tool_execution.py; compaction orphaned-assistant-id sanitization in src/agents/memory/openai_responses_compaction_session.py; corresponding tests added/expanded.
• Files: src/agents/extensions/memory/sqlalchemy_session.py, src/agents/run_internal/tool_execution.py, src/agents/memory/openai_responses_compaction_session.py, tests/extensions/memory/test_sqlalchemy_session.py, tests/test_run_step_execution.py, tests/memory/test_openai_responses_compaction_session.py
• Action: Run uv run pytest -q tests/extensions/memory/test_sqlalchemy_session.py tests/test_run_step_execution.py tests/memory/test_openai_responses_compaction_session.py; pass criteria is stable pass (especially concurrency/cancellation cases).

No material risks identified.

Notes:

• Base tag determined from local tags only, per request: v0.12.5.
• Target set to current HEAD: ac53702473cf5849c8c13ce805d13d9a350666bb.
• Working tree is clean (git status --short returned empty).
• Local verification was not re-run in this review; assessment is based on diff/log evidence and test coverage added in the release range.