Trying out the Absurd queue for AI Workloads

I've been experimenting with Absurd, an experimental durable execution system by Armin Ronacher (the creator of Flask, Jinja, and Rye). It's explicitly not production-ready, but I wanted to see what kind of niche it might fill. My test case: AI workloads with expensive LLM API calls. → Jump into the demo app: absurd.leblancfg.com → Source code for the demo app: leblancfg/absurd-test The Problem with AI Workloads If you're building anything that makes LLM API calls, you know the pain: requests can take anywhere from 2 to 60+ seconds, they occasionally fail, and rate limits are sometimes unpredictable. Traditional task queues like Celery or RQ weren't designed for this. They'll restart your failed task from scratch, re-running all those expensive API calls you already paid for. Durable execution systems like Temporal solve this by checkpointing your work as it progresses. If a worker dies, the next one picks up exactly where you left off. Self-hosting Temporal means running a…