Skip to content

Guides

Practical paths from an empty file to a stage running against a live broker.

Guides

The reference explains what each piece is. These guides show you how to use them — task-focused, worked end to end, and safe to copy from.

Start with Getting started: install the framework, learn the two-yield contract, and run a complete stage against any Kafka broker with a single call. From there the guides branch out by stage shape — Extractors bring an external source into Kafka (with MQTT Extractors as the push-driven variant), and Transformers consume topics and publish derived records with exactly-once delivery. Best practices then shows how to combine an extractor and a transformer so you can reprocess without re-ingesting, and Observability covers the Prometheus metrics. The concept pages fill in the rest: typed attributes and records, config topics as shared lookup tables, and exactly-once delivery.

The Whole Contract Is Two yield Statements

Everything here is built on one idea: a stage is an async generator that yields a Message to emit a record and yields a State to persist state for the current key. No agents, no tables, no DSL. If you keep that shape in mind, every guide below is a variation on it.