Rivers is an orchestration platform for data and ML pipelines, written in Rust for native performance with a Python-first development experience.
Rust core handles planning while Python defines assets, unlike Airflow's Python-heavy scheduler.
Data engineers and ML platform teams
Airflow · Dagster · Prefect
My most recent release adds OIDC and forward auth support. Security shouldn't be paywalled :)
Checkpoint-based recovery beats determinism constraints; embed without separate server, unlike Temporal/Airflow.
Python syntax on Rust Actix core reaches 20k RPS, but Mojo and uv already address Python speed.
Auth layer with zero authority—can't sign or decide, only relay hashed requests.
Lyapunov stability theory for LLM loops beats hard budget caps any day.
Git worktree isolation per agent prevents race conditions in multi-agent coding.
Pydantic models as database models kills the duplicate definition problem SQLModel still has.