Sources

GlassFlow supports two types of data sources. Each source type defines how data enters the pipeline before it is processed and written to ClickHouse.

Source	Description	Best for
Kafka	Consume from one or more Kafka topics with support for deduplication, joins, filtering, and transforms	Event streams, CDC, and any workload already backed by Kafka
OTLP	Receive OpenTelemetry logs, traces, and metrics directly via the OTLP protocol — no Kafka required	Observability data from OTel-instrumented services

Both source types write to ClickHouse through the same sink configuration and support the same transformations (filter, stateless transformation, deduplication). Joins are only available with Kafka sources.