pg_durable

Run `npx skills add microsoft/pg_durable --all -g -y` in your terminal to install the skill globally. The extension itself is added to PostgreSQL via a Debian package, Docker image, or `CREATE EXTENSION pg_durable;`.

It makes long-running database jobs survive crashes and restarts by checkpointing each step, so you never rerun completed work or rebuild state by hand.

For workflows that live mostly in Postgres, yes. It removes the need for external orchestrators, queues, and worker infrastructure by running durable execution as a PostgreSQL extension.

Compose steps with operators like `~>` for sequence, `&` for parallel, and `?>` for conditions, then launch the graph with `df.start()` to get an instance ID you can monitor.

No. It runs entirely as a PostgreSQL extension with a background worker, so there is no Redis, message queue, or separate orchestration service to manage.