Skip to content

chore: release v0.1.21 (#276)

Choose a tag to compare

View all tags
@memtomem memtomem released this 29 Apr 05:17
v0.1.21
82db209
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

chore: release v0.1.21 (#276)

Bumps pyproject.toml + src/memtomem_stm/init.py to 0.1.21, syncs
uv.lock, and stamps the [0.1.21] CHANGELOG section with the four PRs
accumulated since v0.1.20.

Highlights:

  • #272 stm_index_stats MCP tool — INDEX-pipeline observability mirroring
    the surfacing-side counter (per-tool attempts + 4-label outcomes).
    Closes the "is auto-extraction actually writing to LTM" question that
    could only be answered by per-call inspection. Advertised tool count
    goes 11 → 12 under the existing observability flag.

  • #274 Token-equivalent budgets for CJK / non-Latin workloads — opt-in
    chars_per_token (proxy / server / tool cascade) + max_result_tokens
    on server / tool. Ships KO calibration (1.85 chars/token vs EN 4.03)
    empirically measured against tiktoken cl100k_base on a 13-pair EN/KO
    doc corpus. Backward compat preserved — defaults reproduce the prior
    hardcoded 3.5 multiplier.

  • #275 mms init --lang ko — applies the calibrated KO preset (4
    numeric fields across proxy and per-server scopes) in one flag.
    Closes the discoverability gap left by #274. EN preset is intentionally
    a no-op so existing TestInit suite passes without supplying a new
    prompt line. ZH/JA placeholders deliberately omitted until analogous
    measurement corpora land.

  • #273 README coverage section — names the MCP-proxy boundary explicitly
    (Claude Code built-ins, Cursor / Windsurf / Desktop built-ins, and
    sub-agent built-in calls all bypass STM). Pure documentation —
    affects expected-ROI framing relative to the "20–80% savings" headline.

Also fixes the recurring __version__ drift between pyproject.toml
and src/memtomem_stm/__init__.py (was at 0.1.18 — same drift class
that v0.1.12 fixed and that recurred during v0.1.13–0.1.20 release
mechanics). Both now sit at 0.1.21.

Validation: 1787/1787 pass under the full CI filter, ruff check clean
on src + tests.

🤖 Generated with Claude Code

Co-authored-by: pandas-studio pandasdataanalysis@gmail.com
Co-authored-by: Claude Opus 4.7 (1M context) noreply@anthropic.com

Assets 2
Loading