yetanotheraryan/graphify-chokidar: Keeps your graphify knowledge graph fresh without burning tokens.

Keeps your graphify knowledge graph fresh without burning tokens. - yetanotheraryan/graphify-chokidar

Summary

graphify-chokidar is a TypeScript file watcher that keeps a graphify knowledge graph in sync with your codebase without manual rebuilds. It classifies changed files as either AST-parseable source code (instant, no LLM call) or document/media files (prompts before an LLM-backed rebuild). The project is MIT-licensed and is a thin wrapper around the standalone graphify tool.

What it is useful for

It is useful when you want a continuously updated code knowledge graph without rerunning graphify by hand or paying for redundant LLM token usage. Engineers working on polyglot repos, documentation-heavy projects, or AI agent workflows can keep the graph fresh in the background.

Behavior documented in the README:

• AST files include .ts, .tsx, .js, .jsx, .py, .go, .rs, .java, .c, .cpp, .cs, .rb, .swift, .kt, .scala, .php, .lua, .zig, .ex, .vue, .svelte, .dart. Saving one triggers an instant rebuild with no prompt, e.g. ✓ graph rebuilt in 1.4s.
• LLM files include .md, .mdx, .txt, .rst, .pdf, .png, .jpg, .svg, .mp4, .mp3. Saving one prints ⚠ LLM rebuild triggered — these files will consume tokens and asks run graphify? [Y/n].
• Always ignored: graphify-out/, node_modules/, .git/, dist/, __pycache__/.
• If no graph exists yet, saving any file builds one from scratch.
• If a new change arrives mid-build, the previous AST build is cancelled and a fresh one starts; repeated LLM rebuilds are collapsed so you are not spammed with prompts.
• Default debounce is 2000ms, so saving 10 files in quick succession fires one rebuild, not ten.

How engineers can use it

The README documents the following setup:

npm install -g graphify-chokidar
graphify-chokidar

Or run it without installing:

npx graphify-chokidar

graphify itself must already be installed and on your $PATH; the README states this as a requirement but does not document that install step, so refer to the graphify repository if setup is unclear.

Common invocations from the README:

graphify-chokidar ./src
graphify-chokidar --debounce 5000
graphify-chokidar --silent
graphify-chokidar --once   # exit after first rebuild, useful in CI

For AI agent workflows, the project ships a SKILL.md. The README documents installing it for Claude Code by copying the file into ~/.claude/skills/graphify-chokidar/, after which /graphify-chokidar ./src works as a slash command inside a session.

Building from source is documented as git clone, npm install, then npm run build (outputs to dist/).