Ofline
Most workflow libraries make you run something extra. Celery needs Redis. Temporal needs its own server. Gravtory uses the database you already have.
Here's what a saga looks like:
GitHub - vatryok/Gravtory: Production-grade durable workflow engine for Python. Atomic checkpointing, saga compensation, parallel execution, and distributed workers, all powered by your existing database. No message broker required. https://github.com/vatryok/Gravtory
If the process crashes mid-saga, it picks up from the last checkpoint. No data loss, no manual recovery.
View: https://youtu.be/o5m77LiZXNY
vatryok - Overview https://github.com/vatryok