Show HN: WolveKafka – Consume Kafka with a plain HTTP endpoint

WolveKafka (WKS) is an open-source service that lets applications consume Kafka events through ordinary HTTP endpoints, without embedding a Kafka client or running a Kafka Connect cluster. It sends records in batches to a handler written in any language and expects a result for each record. The handler can mark records successful, request a retry, or report failure; WKS then commits offsets, retries with backoff, or applies a configured dead-letter or drop policy. A handler can also return HTTP 429 with Retry-After to pause delivery without using up retries. WKS supports routing batches among weighted HTTP targets, ordered or parallel processing, and configuration reloads without restarting.

The project includes a Docker Compose demo with Kafka and sample Python handlers, plus exercises for retry recovery, dead-letter delivery, and routing. For operations, it offers Prometheus metrics, Grafana dashboards, OpenTelemetry tracing, and SASL, TLS, and mutual TLS support, with Helm guidance for Kubernetes deployment. Its delivery guarantee is at least once: crashes may produce duplicates, so handlers should make side effects idempotent. In parallel mode, records within a partition may be reordered. WKS is aimed at teams whose business logic belongs in separately deployed services and who want shared Kafka-consumer plumbing; a direct consumer may be preferable when an application needs finer client control, while connector or cloud-managed alternatives may suit teams already committed to those platforms.