Reaparr

A read-only dashboard that surfaces stale, unwatched media on your *arr stack as priority-ranked deletion candidates — and shows how much space you'd reclaim.

It consolidates four sources into one view per title:

Source	Provides
Sonarr	Series structure + size on disk
Radarr	Movie files + size on disk
Seerr	Who requested what, and when
Tautulli	Who actually watched what, and when

Each title gets an opinionated Reap Score (0–100, higher = stronger candidate), with size shown beside it as the payoff — never folded into the score. People are resolved into a single canonical person across Seerr and Tautulli automatically.

MVP is read-only: it ranks and tracks, it does not delete. No auth (trusted LAN).

Quick start

npm install                 # builds the better-sqlite3 native module
REAPARR_DEMO=1 npm run dev   # boot fully populated with demo data, no sources needed

Open the printed URL. The dashboard renders ranked Series / Movies, the People page shows identity reconciliation, and Settings → Connections configures sources.

Demo mode

REAPARR_DEMO=1 seeds a realistic fake picture (6 series, 4 movies, requests, watch history, pre-mapped + conflicting people) so every view works end-to-end with zero configuration. Sync now refreshes it.

Mock source server

Exercises the real adapter + sync code paths (all four auth shapes) with no credentials:

npm run mock-server          # serves the demo dataset at each source's real API paths

Then in Settings, point each connection's Base URL at http://localhost:8900 with any non-empty key, Test, enable, and Sync.

The Reap Score (v1)

raw = Staleness(0–50) + Abandonment(0–30) + Request-miss(0|20), then a grace factor ramps brand-new content in over ~30 days. Driven by watch + request signals only:

• Staleness — stepped at 3 / 6 / 12 months idle (Stale → Very stale → Dormant).
• Abandonment — 30 × (1 − completion); series completion = episodes watched ÷ on disk.
• Request-miss — requested but never watched by anyone → +20.

Size is never in the score (it's an independently-sortable companion column). The only two knobs (Settings → Scoring): the staleness tiers and the grace period.

Tests & validation

npm test          # 36 unit + integration tests (scoring, identity, join, full pipeline, mock sync)
npm run typecheck
npm run build

The build is self-validating (spec §12.5): unit tests lock the scoring formula and identity rules, an end-to-end test runs the demo dataset through the full pipeline with known expected scores, and another runs a complete sync against the mock server. Only the final "enter real API keys + one live sync" step needs you.

Deployment (Docker on Unraid)

docker build -t reaparr .
docker run -d -p 3000:3000 -v /mnt/user/appdata/reaparr:/app/data reaparr

The SQLite database lives in the mounted /app/data volume and persists across updates. Configure sources via the Settings UI (or seed with SONARR_URL/SONARR_API_KEY/… env vars).

Architecture

• Nuxt 4 / Vue 3 + NuxtUI, Nitro node-server preset.
• SQLite via Drizzle ORM + better-sqlite3, on a volume-mounted file.
• Daily sync (Nitro scheduled task) + manual Sync now; the dashboard reads the local cache, never the live sources.
• Pure transforms (server/sync/score.ts, identity.ts, join.ts) are fully unit-tested.

See docs/reaparr-spec.md and docs/reaparr-technical-plan.md.