> For the complete documentation index, see [llms.txt](https://docs.hardpoint.dev/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.hardpoint.dev/guides/invar/integrations.md).

# Integrations

Compatibility notes and getting-started snippets for specific projects and libraries

### BullMQ

[BullMQ](https://bullmq.io/) is a popular job queue built on Redis, and one of the projects Invar's Redis compatibility is actively tested against.

#### Connecting

BullMQ needs one non-default connection option to work against any Redis-compatible target, including Invar: `maxRetriesPerRequest: null`, which BullMQ's own documentation requires for `Worker` connections, since its default retry-limit-per-request is incompatible with BullMQ's blocking-read pattern:

```typescript
import { Queue, Worker } from 'bullmq';

const connection = {
  host: 'localhost',
  port: 6379,
  maxRetriesPerRequest: null,
};

const queue = new Queue('my-queue', { connection });

const worker = new Worker('my-queue', async (job) => {
  // process job.data
}, { connection });
```

Beyond that, it's the same BullMQ setup you'd use against Redis — swap `host`/`port` for wherever your Invar instance is running and everything else is unchanged. See Quick Start if you don't have an instance running yet.

#### Test coverage

Invar passes BullMQ's own smoke test suite in full, and roughly 90% of BullMQ's broader test suite, (as of `vpy3.2.0)`. The remaining gap mostly traces back to a small number of not-yet-implemented commands rather than anything BullMQ-specific — consumer group support on streams (`XACK`, `XCLAIM`, `XGROUP`, `XREADGROUP`, and related) is the most likely source, since BullMQ's event stream is a real stream underneath, and that whole command family isn't implemented yet. See the Redis [command support appendix](https://docs.hardpoint.dev/guides/invar/appendix/redis-tm-command-support.md) for the authoritative, up-to-date list of what's implemented.

#### Conflict behavior under heavy contention

Invar uses optimistic concurrency control under the hood — concurrent transactions that touch overlapping keys can conflict, and Invar automatically retries a conflicting transaction with jittered backoff before it's ever surfaced to the client. For the vast majority of workloads this is entirely invisible; you won't see it or need to think about it.

Where it becomes relevant: BullMQ's own queue-management keys (`wait`, `active`, `marker`, and similar) are, by design, high-contention — every worker pulling from the same queue is writing to the same small set of keys. Under sustained, heavy load with many concurrent workers on one queue, this can mean more retries and correspondingly higher tail latency on job claims than you'd see with real Redis's single-threaded execution model, and in extreme, sustained contention a command can exhaust its retry budget and fail outright rather than eventually succeeding.

This is a known, active area of tuning rather than a hidden gotcha — if you're running many workers against one queue at high throughput, it's worth being aware this exists rather than being surprised if you see it under load testing. We'd love to hear your experience with Invar under these conditions- either create a [GitHub issue](https://github.com/hardpointlabs/invar/issues) or ping us on [Discord](https://discord.gg/WWE4PWVnb2)!
