@loomfsm/repo-memory

Persistent, model-agnostic structural memory of a code repository. It maintains a plain-markdown repo brief — key types, public API, layout, and stack, each with a file:line anchor — so a planning agent cites structure from the brief instead of cold-reading the whole tree on every run. The brief persists across runs under loom's footprint and delta-refreshes cheaply, so the second task on the same project starts from a warm map rather than re-deriving it.

What's inside

• ensureBrief(projectDir) — builds or delta-refreshes the brief at .loom/memory/<hash>/repo-brief.md. Lists tracked files, content-hashes them, and re-extracts only the files that changed (a content-hash cache carries the rest verbatim); an unchanged tree is reused byte-for-byte. Degrades to a no-op (never throws) on a non-git project, a repo past the size cap, or any error — a stale brief can only cost tokens, never correctness.
• Importance ranking — the brief orders files by module in-degree (how many other files import them), then public-surface size, fitting the highest-signal contracts into a token budget instead of dumping the tree.
• Cross-model by construction — the output is plain markdown and JSON. Any model or backend consumes it identically; a cheap pass can author what an expensive planner reads.

It is ambient: no clock or replay state is threaded through it, so it lives entirely outside loom's deterministic kernel — which never depends on this package.

Part of loom

loom drives multi-step LLM agent work — code review, implementation, any review-gated task — as a replay-deterministic state machine: safety invariants enforced at commit time, human gates where they matter, and a complete, replayable audit trail in a local SQLite file.

Most users should install @loomfsm/pipeline (npm i -g @loomfsm/pipeline), which pulls the whole runtime in one step. Install this package directly only if you are assembling your own runtime.

Website · Quickstart · Why loom · GitHub

License

Apache-2.0