Temporal, Airflow, Step Functions are great — but sometimes you just need "run these steps, resume where it stopped if something crashes," without the infrastructure and learning curve that comes with a full platform, and whole mental model to internalize

So I built something simpler and maybe i over simplified things..

It sticks with philosophy of "workflows as code", so no dsl , no no-code, or visual designer. just a developer friendly lib..

Sayiir is a library you pip/npm/cargo install. Define steps as normal async functions, it checkpoints each output. On failure, completed steps are skipped, execution picks up where it left off.

No replay, no determinism rules, no separate server to deploy, no activities — just your good old existing async functions with durable execution underneath.

Rust core, Python/Node.js bindings. MIT licensed

Docs: https://docs.sayiir.dev Playground: https://docs.sayiir.dev/playground/