graphifylisted

# graphify: query a code graph instead of re-reading files graphify (CLI, already installed via `uv tool install graphifyy`) parses a repo with tree-sitter into a graph of nodes (functions, classes, files, concepts) and edges (`calls`, `imports`, `implements`, ...), each tagged `EXTRACTED` (explicit in source) or `INFERRED` (derived). Build the graph once, then navigate it with cheap traversal queries instead of dumping whole files into context on every question. Use this instead of repeated grep+read cycles when: exploring an unfamiliar codebase, answering "how does X connect to Y", finding every caller of a function, or tracing a dependency chain across files. ## Workflow ### 1. Build the graph once ```bash graphify update <repo-path> --no-cluster # fast: skip community clustering graphify update <repo-path> # with clustering (adds community labels to the report) ``` This is the build step - despite the name, `update` does the initial build *and* every incremental refresh afterward (only changed files are re-extracted on repeat runs). Code extraction is pure tree-sitter AST: no LLM, no API key, nothing leaves the machine. Output lands at the default location: ``` <repo-path>/graphify-out/graph.json ``` Pass `--force` if a later `update` needs to shrink the graph (e.g. after deleting a lot of code) - graphify otherwise refuses to overwrite a bigger graph with a smaller one. Details, plus `cluster-only` and `watch`: `references/update-and-watch.md`.