[gloryfromca]

Summary

Stable 1.5.0. The Unreleased entries move under ## [1.5.0] - 2026-10-21 with the release page's lead paragraph, and Upgrade gains one note (the first start creates the memorize_inbox table; async_mode is opt-in). Version bump in pyproject.toml, docs/openapi.json (regenerated after uv sync) and the project's line in uv.lock. The code is the same as 1.5.0rc1 plus the release workflow fix in #526.

Area

• Architecture method
• Benchmark
• Use case
• Documentation
• Developer experience
• CI, build, or release

Verification

make lint                                   # incl. dump_openapi --check: OK
uv run pytest tests/unit -k "app_metadata or version"   # 11 passed
release.yml notes script, run locally for v1.5.0 on this branch:
  finds the section; page = lead paragraph, Added / Changed / Fixed, Upgrade (stable pip line); prerelease=false

On 1.5.0rc1 (same code): full CI incl. slow / live_llm / e2e on macOS, Windows integration, smoke with recall and a search-relevance read-through on macOS / Windows, a PyPI-resolved install smoke of the published rc, hypothesis deep on every state machine, a 2 h queue soak with fault injection (0 reconcile failures), load + idle soaks on macOS (2 h), Windows (2 h) and Linux (4 h) with zero request errors, and an upgrade check: the PyPI 1.5.0rc1 server started on a root built by 1.4.2 (inbox table created, old data searchable, async /add 202 then success, embedding model recorded, no traceback).

Known and accepted for this release: on Windows, search latency grows over a long run (present in 1.4.2, reduced here); Linux RSS under sustained writes has no observed ceiling in 1.4.2 or this release and is under investigation.

Checklist

• I kept the change scoped to the relevant area.
• I am opening this from a separate branch, not pushing directly to main.
• I updated docs, examples, or setup notes when behavior changed. (CHANGELOG, docs/openapi.json)
• I added or updated tests when the change affects behavior. (no behavior change)
• I did not commit secrets, .env files, dependency folders, or generated output.
• Active relative links in Markdown files resolve.

Notes for Reviewers

After merge: tag v1.5.0 on main and approve the release environment; the workflow lifts the section above into the GitHub Release page.

🤖 Generated with Claude Code

[gloryfromca]

No blockers; this can merge as far as I am concerned.

Reviewed the actual PR delta (0595134...08c94f1) and the requested local origin/main...HEAD comparison, including the precursor commits included by the stale local base. The 1.5.0 version is consistent across pyproject.toml, uv.lock, installed package metadata, and generated OpenAPI; the stable changelog section matches the release workflow's parser and documents the new queue startup effect.

Coverage included CLAUDE.md, the release procedure and engineering rules; the diff and relevant release history; metadata/release-workflow consumers; backward compatibility; test-weakening review (none in the actual PR delta); and architecture constraints (no layer dependency impact).

Verification: local unit suite 2,915 passed / 4 skipped; local integration suite 188 passed / 9 skipped / 7 deselected; OpenAPI drift check passed. The documented aggregate make ci could not start because this environment lacks make, so its test targets were invoked directly; hosted lint, package build, links, commit-message, title, and Linux integration checks also pass.

[gloryfromca]

No blockers; this can merge as far as I am concerned.

Reviewed the new 08c94f1...1d1d9ae delta and rechecked it in the full PR comparison. The added notes accurately reflect the merged #492 per-worker handler bound and #490 projected/skip-vector reads, and both land in the generated release page's Fixed section.

This round covered the repository and release rules, the delta and underlying implementation/history, release-note extraction callers, backward compatibility, test weakening (none; this delta only changes CHANGELOG.md), and architecture constraints (no code or dependency-boundary change). Documentation checks, contributor-doc checks, release-note extraction, and git diff --check pass; hosted lint, package build, links, commit-message, and title checks also pass, with the remaining hosted test matrices still running. No prior review threads exist to sweep.