Skip to main content

Formats

A format (deserializer) decodes raw source payloads into typed records before they reach the operator chain. It is optional — a source that already emits ready-made records needs none. Configure one with a single-key deserializer: section in the pipeline YAML.

FormatConfig tagSummary
Avrodeserializer: { avro: ... }Confluent/schema-registry and single-object Avro; non-blocking schema fetches, Skip/Fail on undecodable payloads, optional fast backend.
JSONdeserializer: { json: ... }serde_json-backed single-document, NDJSON, and top-level-array framings; typed or dynamic records, per-record Skip/Fail with etl_json_deser_* metrics.

Deserialization is a small trait — the same Extending guide covers writing your own decoder.