Skip to content

Second Prerelease - Analyze large ts codebase identity visually

Choose a tag to compare

@finnmglas finnmglas released this 08 Aug 17:14
· 157 commits to main since this release

Run npm i desprawl, then desprawl view in any git repo. Analyses git-tracked files.

✨ Opus 5 Summary

Project metadata

  • what the repo is, read off manifests and marker files: language and strictness,
    package manager, pinning, frameworks, state, ui, what it connects to, auth,
    telemetry, ports, build, testing, lint, format, ci, containers, api collections
  • every package.json in the tree, so a monorepo is read whole
  • identity above it: name, version, licence, and what it holds (frontend, backend,
    cli, monorepo, mobile, desktop, infra)

Assistance

  • which ai coding tools left a trace, from checked in rules (CLAUDE.md, AGENTS.md,
    .cursor, .aider, .windsurfrules and twenty more) and from commit signatures
  • Co-authored-by, Assisted-by and Generated with trailers, plus bot author
    addresses. Prose is never matched, so a commit about an assistant is not one by one
  • how much of the read history each tool signed

Explorer

  • desprawl view now serves live and reanalyses on request, so refresh picks up new
    commits without a restart. One fixed origin, so display settings persist between runs
  • drag across the time series to zoom it, double click or reset to go back
  • click a commit in the history graph for its files, added and removed
  • time series with selectable granularity, a log scale, and per-series toggles
  • every column explains itself on hover, every table copies as tsv or csv

Scale

  • runs on the linux kernel: 87k files, 1.4m commits. The log reads a capped window
    and says so, while commits, contributors and repo size over all time come from
    cheaper passes

No runtime dependencies.

https://www.npmjs.com/package/desprawl