feat(knowledge): typed memory schema — kind, subject, review_state, expires_at (#3072)

[mabry1985]

Summary

• Adds 4 typed memory columns (memory_kind, subject, review_state, expires_at) to the knowledge store via additive SQLite migration — no delivery behavior changes
• Adds memory_kind and review_state filter parameters to search(), list_chunks(), and corresponding memory tools (memory_recall, memory_list)
• Updates memory_ingest tool to accept memory_kind and subject, and stamps session namespace on writes

Tracks #3072. Additive migration: all columns are nullable with None defaults; existing callers are unaffected. The Chunk dataclass, as_dict(), and all tool output formats are extended — never broken.

Test plan

• ruff check . passes
• python -m pytest tests/test_typed_memory.py -q — 14 tests covering migration, round-trip, filters, backward compat
• python -m pytest tests/test_knowledge_lifecycle.py -q — existing lifecycle tests still pass (37 total)
• CI gates (ruff, lint-imports, full pytest)

🤖 Generated with Claude Code

https://claude.ai/code/session_01WEMxBi71vjtmmmziFCMcby

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[mabry1985]

Duplicate — work consolidated on #3205.