SnatchArr¶
Keeps your *arr libraries complete without hammering your indexers. Gentle by default, cloud-native from day one, and yes, the name means exactly what you think it means.
SnatchArr asks Sonarr, Radarr, Lidarr, Whisparr (v2 and v3) and Readarr (legacy) to search for missing items and cutoff-unmet upgrades in small batches. Every search is a snatch. It remembers what it already snatched (the afterglow), never exceeds an instance's hourly stamina, and slows down (foreplay) when your download client is busy, paused or gone.
It is a from-scratch rewrite of the Huntarr lineage (newtarr): same intent, none of the code.
Why another one¶
| Huntarr / newtarr | SnatchArr |
|---|---|
| One setting set per app, shared by all instances | A snatch policy per instance |
| Hourly cap checked once per cycle | Stamina debited per dispatch, shared across workers |
| Processed items never expire | Afterglow with a TTL per item, separate for missing and upgrades |
| Looks at the *arr queue size only | Watches qBittorrent, Transmission, Deluge, rTorrent, SABnzbd and NZBGet for backpressure |
| Python, single process, JSON files | Go control plane, Rust worker, Postgres, Helm, OpenAPI 3.1 |
Where things live¶
- Concepts: snatches, afterglow, stamina, foreplay, quickies.
- Configuration: environment, profiles, Configarr link.
- Download clients: discovery, backpressure, bandwidth pacing.
- Seerr: requests snatched first, request dashboard, instance import.
- Prowlarr: indexer stamina, limits, and gate semantics.
- API: the OpenAPI contract and how to talk to it.
- Deployment: images, Helm chart, CloudNativePG, ArgoCD, kind.
- Architecture decisions: why it is built the way it is.
Status¶
Pre-alpha, released: versions and their notes are on the releases page.
Code is EUPL-1.2, documentation is CC BY-SA 4.0.