Inngest vs RQ (Redis Queue)

Inngest runs asynchronous code as durable functions for event-driven workflows, background jobs, and agents. Its execution steps handle retries, retain state, and recover from failures; traces show activity from trigger to completion. Flow controls manage concurrency and fairness when users compete for resources. Developers can use open-source SDKs for TypeScript, Python, and Go, or the REST API. Functions can run wherever an app is hosted, with deployment guides for services including Vercel, Netlify, Cloudflare Pages, Render, and DigitalOcean. Inngest also describes isolated sandboxes for testing changes and tools for scoring and comparing variants on production traffic. The Free plan includes 5 seats, 50k executions, 500k events ingested, and 24-hour trace history. Execution pauses when the free quota is reached until the account upgrades. Pro starts at $99/mo. Inngest states that it is SOC 2 Type II compliant, and that its US-based databases are encrypted at rest and application traffic uses TLS/SSL in transit.

RQ (Redis Queue) is a Python library for placing jobs in queues and running them asynchronously in background worker processes. It is self-hosted, open source under the BSD license, and free. It requires Redis 5 or later, or Valkey 7.2 or later. RQ supports scheduled and repeating jobs, retries, bulk enqueueing, job groups, dependencies, unique jobs, and concurrency rate limits. Unique jobs cannot be combined with dependencies. Failed jobs are recorded with exception information, and custom exception handlers are supported. For monitoring, there is an `rq info` console command; RQ Dashboard is a separate web-based frontend. Documentation covers Django, including django-rq, as well as usage with Heroku, Sentry, Supervisor, and systemd. RQ serializes jobs with pickle by default, which the project warns is safe only with trusted Redis instances; JSON is an alternative. Workers need a system that supports `fork()`, so Windows requires Windows Subsystem for Linux and a bash shell. Workers must also be able to import job functions, and producers and workers must share source code.