// the find
Patel-world/react-native-durable-queue
Durable SQLite-backed job queue for React Native — retries with backoff, dead-lettering, idempotency keys. C++ core over JSI.
A durable job queue for React Native that stores jobs in a vendored SQLite database behind a C++ core over JSI. It offers retries with jittered exponential backoff, dead-lettering, idempotency keys, priorities, and delayed jobs, so uploads and sync calls survive app restarts. It is for apps that need reliable retries while the app is in the foreground, which is the only mode it supports today.
Active jobs carry a leased_until timestamp instead of a boolean lock, so a job whose worker was killed is picked up on the next dequeue rather than stranded, which is where most hand-rolled queues fail. The jitter argument is specific to phones: many devices retrying on the same network-restored event is a real stampede, and the README explains why it matters there. Attempts count leased runs rather than only failures, so a job that reliably crashes the app dead-letters instead of boot-looping users. The core is plain C++ tested on the host in seconds, with 43 checks that include lease reclamation after a simulated crash and persistence across close and reopen.
Nothing runs while the app is backgrounded, and background draining is only on the v0.2 roadmap, so the current version is a persistent foreground queue. The README does not say whether a handler that outlives leaseMs can be handed out again; if it can, the job runs twice, so handlers must be idempotent even with idempotency keys, and that should be stated in the semantics section. The tests shown are host-side C++ against SQLite, and I could not see tests for the JSI bridge or the TypeScript Queue wrapper in the visible tree, which is truncated, so the JS-to-native seam looks unverified. The project is brand new with zero stars and requires the New Architecture, which narrows who can adopt it.