GitHub only retains 14 days of clone-traffic data, and there's no
third-party archive of historic clones. So we ship our own tracker:
- `scripts/track-clones.mjs` — fetches clones for 5 ruflo-ecosystem
repos (ruflo, agentdb, agentic-flow, ruvector, ruv-FANN) and npm
downloads for headline packages. Produces three artifacts:
- `data/clone-data.rvf` — RuVector vector store (RvfDatabase,
10-dim per snapshot for similarity search across time)
- `data/clone-data.ledger.json` — append-only chronology with
per-repo breakdown (human-readable in PR diffs)
- `data/clone-data.proof.json` — provenance manifest with
SHA-256 over the ledger
- `.github/workflows/clone-tracker.yml` — runs every 13 days at
06:17 UTC + on workflow_dispatch + on edits to the tracker.
Commits the new snapshot back to main as `ruflo-bot`.
README badges now show:
- `ecosystem downloads 22.2M+` (sum across claude-flow + ruflo +
@claude-flow/* + agentdb + agentic-flow + ruvector/* — 152 npm
packages, last 12 months: 22,275,946 as of 2026-05-19)
- `git clones 14d 115k` (linked to the live ledger)
Initial snapshot captured at run time. Sources/formula documented
in `data/clone-data.proof.json` so any reader can verify the
claim by re-running the script.
Co-Authored-By: RuFlo <ruv@ruv.net>
ruv