// the find
warpstreamlabs/bento
Fancy stream processing made operationally mundane. This repository is a fork of the original project before the license was changed.
Bento is a declarative stream processor — you write YAML configs to wire sources, processors, and sinks together without writing application code. It's a community fork of Benthos, created after the original project relicensed under a commercial-unfriendly license in 2024. Aimed at backend engineers who want Kafka-style data pipelines without the Flink/Spark overhead.
Bloblang is genuinely well-designed — a purpose-built mapping language that handles JSON reshaping, type coercion, and conditional logic cleanly without dropping into code. The connector list is legitimately wide: AWS, GCP, Azure, Kafka, NATS, Redis, SQL, MQTT, AMQP, and more, with integrations tested in CI. The in-process transaction model gives you at-least-once delivery without a separate state store — crash recovery just works. Plugin API is a real first-class citizen; you can extend inputs, outputs, and processors in Go without forking the core.
Being a fork means you're betting the project survives without the original team — contributor velocity here is much lower than the original Benthos had before acquisition, and the fork is only ~2K stars. The release process is explicitly unfinished ('we're working on it') and there are no pre-built binaries linked from the README, which is a real friction point for evaluators. Documentation lives on a GitHub Pages site that mirrors the original Benthos docs with varying degrees of accuracy; some pages will describe behavior from the pre-fork codebase. No native schema registry integration for Kafka, which is a gap if you're in a serious Kafka environment.