Release 0.14.6

[github-actions]

Release readiness review (v0.14.5 -> TARGET 874b137)

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

Diff

v0.14.5...874b137

Release call:

🟢 GREEN LIGHT TO SHIP No release-blocking regression, packaging break, data-loss, security, or unversioned breaking API evidence was found.

Scope summary:

• 104 files changed (+2644/-2466); key areas touched: GPT-5.5 model refresh in docs/examples/runtime defaults, MongoDB session docs, CI uv pinning, dependency resolver settings, websockets upper-bound relaxation, and focused tests.
• Commit range reviewed: 7 commits from c2cb0314 chore: harden uv dependency resolution (#3014) through 874b1377 Bump version to 0.14.6.

Risk assessment (ordered by impact):

No material risks identified.

1. GPT-5.5 model refresh across runtime defaults, examples, and docs

   • Risk: 🟢 LOW. The most visible user-facing change is defaulting many examples and sandbox memory phase-two consolidation to gpt-5.5, but the SDK's base default model remains documented as gpt-4.1 and callers can still override models.
   • Evidence: src/agents/sandbox/config.py changes MemoryGenerateConfig.phase_two_model from gpt-5.4 to gpt-5.5; src/agents/models/default_models.py, src/agents/models/openai_responses.py, and tests add gpt-5.5 handling for default reasoning and GA computer-tool payload selection.
   • Files: src/agents/sandbox/config.py, src/agents/models/default_models.py, src/agents/models/openai_responses.py, tests/models/test_default_models.py, tests/test_openai_responses_converter.py, tests/sandbox/test_compaction.py, docs/models/index.md, docs/tools.md, examples/**
   • Action: Run uv run pytest tests/models/test_default_models.py tests/test_openai_responses_converter.py tests/sandbox/test_compaction.py -q; pass criteria: all focused model-default, computer-tool conversion, and compaction context-window tests pass.

2. Dependency and CI resolver changes

   • Risk: 🟢 LOW. The release/build path now pins astral-sh/setup-uv to a newer commit with uv 0.11.7, adds exclude-newer resolver settings, and relaxes websockets from <16 to <17; this is not a confirmed break, but it touches packaging and CI behavior.
   • Evidence: pyproject.toml adds [tool.uv] and [tool.uv.pip] settings and changes websockets>=15.0,<16 to <17; uv.lock records the new resolver options and dependency specifiers; workflows pin setup-uv with version: "0.11.7".
   • Files: pyproject.toml, uv.lock, .github/workflows/docs.yml, .github/workflows/publish.yml, .github/workflows/release-pr.yml, .github/workflows/tests.yml, .github/workflows/update-docs.yml
   • Action: Run the release packaging path with uv 0.11.7: make sync && uv build; pass criteria: dependency sync completes from the lockfile and both sdist/wheel artifacts build successfully.

3. MongoDB session documentation surfaced in docs navigation

   • Risk: 🟢 LOW. This adds documentation for an existing memory extension and navigation entry, so the main risk is docs build/reference drift rather than runtime behavior.
   • Evidence: New docs/ref/extensions/memory/mongodb_session.md, new MongoDB session section in docs/sessions/index.md, and mkdocs.yml navigation entry; no matching runtime code change in this diff.
   • Files: docs/ref/extensions/memory/mongodb_session.md, docs/sessions/index.md, mkdocs.yml
   • Action: Run make build-docs; pass criteria: MkDocs builds without unresolved MongoDBSession reference warnings or navigation errors.

Notes:

• Latest local release tag was determined with local tags only: v0.14.5.
• TARGET is current HEAD: 874b137742d6bbb38cb5da09542556fe7f7d0da7.
• Working tree was clean during review.
• git diff --check v0.14.5...874b137742d6bbb38cb5da09542556fe7f7d0da7 produced no whitespace/conflict-marker output.
• Local verification commands were not run because uv was not installed in the CI shell used for this report; this is not a release blocker under the review policy.