DATABASES Signal 477
Engrim ships local SQLite memory engine letting developers switch AI CLI tools mid-project without losing context
Engrim is a local-first SQLite memory engine that externalizes project decisions, constraints, and state so developers can switch between AI coding agents like Claude Code, Cursor, and Google Antigravity on the same project without losing accumulated context.
As context windows scale, attention dilution degrades reasoning and multiplies cost on every conversational turn, while clearing context causes total amnesia across model switches. Engrim consolidates large amounts of session work into a small curated memory pack that reloads intact on session restart, claiming a 99%+ cut in reloaded context cost. This is a single Show HN post with no independent corroboration, so the claims rest solely on the author's self-reported case study.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Engrim combines SQLite FTS5 full-text search with static vector embeddings (model2vec) in a hybrid reciprocal-rank fusion retrieval engine stored locally at ~/.engrim/memory.db.
A self-reported 105-session case study on a 50,000-line algorithmic trading system claims zero regressions across 186 unit tests and zero context amnesia across model switches.
Installation is via pip install engrim, with auto-detection for installed environments or explicit per-platform setup commands for Google Antigravity, Claude Code, Cursor, and Windsurf.
THE CLUSTER