wpaalperen/immich-revive

Immich Revive is a small background program that keeps Immich, a self-hosted photo and video management app, running smoothly when its machine learning processing fails. Immich can offload tasks like face detection or smart search to a separate remote machine, but if that machine goes offline partway through, those jobs fail and Immich does not retry them automatically. Normally someone would have to log into the admin panel and restart the failed jobs by hand every time this happens. This tool solves that by running continuously alongside your Immich server and checking the Immich API on a regular interval, every two minutes by default. When it finds jobs that have failed in the queues you choose to watch, such as face detection, smart search, or metadata extraction, it clears those failed entries so Immich treats the affected photos as unprocessed again, then triggers reprocessing. The result is that as soon as the remote machine learning server comes back online, those photos get picked up and finished automatically, without anyone needing to intervene. The daemon works with both older and newer versions of Immich, automatically detecting which version of the job API your server uses. It can be run either as a Docker container alongside your existing Immich setup or as a systemd service directly on the host machine, and it needs an administrator API key from your Immich instance to function. Configuration, including which job queues to monitor and how often to check, is handled through a simple environment file. This project is aimed at people who already self-host Immich with a separate remote machine learning server, particularly setups where that ML machine is not always online, such as a laptop that only sometimes joins the network. It is released under the MIT license.