v1.1.0

1.1.0 (2026-08-13)

Features

• memgw key off to turn extraction off, and skip worker/notes entirely when no LLM key (1f273e6)

v1.0.6

1.0.6 (2026-08-11)

Miscellaneous Chores

• automate releases with release-please (802cd39)

memgw v1.0.4

[1.0.4] - 2026-08-10

Fixed

• memgw start while a gateway is already running printed a raw EADDRINUSE
  stack trace; it now says "memgw is already running -- nothing to start" and
  exits 0 (pre-flight health check before binding). A port taken by something
  that is NOT memgw gets a one-line explanation instead of a crash dump.

Added

• setup now installs the Codex transcript watcher as a supervised process
  (launchd com.memgw.watch-codex on macOS, systemd --user
  memgw-watch-codex on Linux) whenever Codex CLI is found and memgw runs
  from a permanent install -- Codex capture no longer needs a manual step.
  Windows / npx-cache runs still get the exact command printed.
• memgw key: add or replace the LLM key after setup with one command --
  masked prompt, provider auto-detected from the key prefix (OpenAI, Anthropic,
  Groq, OpenRouter -> endpoint and default model configured automatically), and
  the gateway is restarted so the key is live. Queued keyless events are
  distilled on the next worker pass.
• memgw status now shows per-agent last-seen times ("agents seen"), so
  "which agents are actually connected" has an answer; /stats by_source gained
  last_ts.

Changed

• setup's key prompt explains that skipping is safe (events queue, nothing is
  lost) and that the key stays on the machine; README quick start now leads
  with the permanent install (npm install -g) and keeps npx as the try-it path.

[1.0.3] - 2026-08-10

Fixed

• doctor could report "gateway not running" while the gateway was healthy
  (stale keep-alive socket or a busy machine timing out the first probe when
  doctor runs inside setup). Both gateway checks now retry once before
  concluding anything.

[1.0.2] - 2026-08-10

Fixed

• The gateway that setup starts directly (when supervision is unavailable) now
  logs to ~/.memgw/gateway.log instead of discarding output, and doctor
  prints the log tail when the gateway is down -- a silently dying gateway was
  undebuggable.

[1.0.1] - 2026-08-10

Fixed

• memgw setup crashed on Node versions where readline's internal
  _writeToOutput is missing (seen on v21.7): masked key input now falls back
  to visible input instead of dying.
• setup and doctor now warn on odd (non-LTS) Node versions, which have no
  better-sqlite3 prebuilds.

Added

• AGENTS.md: an agent playbook for installing memgw on a user's machine and
  for working on the codebase (shipped in the npm package).

memgw v1.0.0

Changelog

All notable changes to this project are documented here.
Format follows Keep a Changelog;
versioning follows Semantic Versioning.

[1.0.0] - 2026-08-10

First public release.

Core

• npx memgw start: local-first, zero-config startup. Creates ~/.memgw, generates
  keys, binds to loopback. The server refuses to start without an auth key, and
  requires a strong key when bound beyond loopback.
• Three-layer store: raw events (90-day retention), distilled facts (forever),
  Markdown topic notes in a git repo (forever). Configuration is centralised in
  src/config.js with a documented precedence chain.
• CLI: setup (one-command wizard: config, supervision, wires every agent found
  on the machine), start, init, status, doctor, search, save,
  forget, embed, watch, hooks.
• Multi-agent capture. Native paths for Claude Code (hooks) and opencode (plugin),
  transcript watcher for Codex CLI and any other JSONL-writing CLI, plus a generic
  parser for unknown formats. MEMGW_CAPTURE_IGNORE excludes directories from
  capture (chiefly the memgw checkout itself).
• MCP server (Streamable HTTP, stateless) with five tools, reachable by header auth
  or by path secret for web clients that cannot send headers.

Retrieval

• SQLite FTS5 with porter stemming and diacritic-insensitive matching.
• Optional semantic layer: hybrid BM25 + vector retrieval with reciprocal-rank
  fusion. Vectors are Float32 BLOBs in the same SQLite file — no vector database,
  no new dependencies. Toggle with memgw embed on|off|status; off by default,
  and search falls back to BM25 whenever the embeddings API is unavailable.
• Serve-time budget caps on /bootstrap (profile size, topic count).

Extraction quality

• Prompts in English and Vietnamese (MEMGW_PROMPT_LANG). Relative time
  ("yesterday") is resolved to absolute dates; assistant recaps of already-stored
  facts are ignored; dedup defaults to "skip" for paraphrases and forbids lossy
  merges (guarded by effectiveness test T4b — a real-LLM suite run before each release, not part of the keyless CI).
• Reasoning models (gpt-5 family, o-series) supported: they receive
  max_completion_tokens and no explicit temperature.

Benchmarks

• scripts/bench-locomo.mjs and scripts/bench-personamem.mjs: end-to-end recall
  benchmarks through the real pipeline, with resume logs. Measured results in
  docs/06-BENCHMARKS.md: LoCoMo 58.6% (BM25) / 66.4% (hybrid), PersonaMem 32k
  59.6%.

Operations

• Retention with three safety rails: only processed events are eligible, the newest
  200 are always kept, and windows under 7 days are refused.
• Backups: Litestream to S3-compatible storage, and an optional git push of the
  notes directory.
• deploy/com.memgw.plist (macOS launchd) and deploy/memgw.service (systemd) keep
  the gateway alive; the bootstrap hook announces a dead gateway in the session
  context instead of failing silently.
• test/verify-docs.mjs fails the suite when documentation drifts from code.