Orchard is an early preview — connector details below are illustrative while we iterate.

Streaming

Kafka

Attach a Kafka cluster and consume topics as DuckDB tables — no Connect cluster, no Flink job, no bespoke consumer to babysit.

Stable By Query.Farm Rust worker

Pricing

What Kafka does

The Kafka connector turns any Kafka-compatible cluster (Apache Kafka, Redpanda, MSK, Confluent Cloud) into an attachable SQL catalog. Topics show up as tables you can SELECT from, filter, and join against the rest of your data.

Consume from an offset or tail the live tail of a topic, decode JSON or Avro payloads into typed columns, and write rows back to a topic with a table function. The worker manages consumer groups, schema-registry lookups, and backpressure so your SQL stays simple.

Highlights

Consume topics as tables with JSON or Avro decoding
Tail live messages or replay from a specific offset
Schema Registry integration for typed columns
Produce rows back to a topic from SQL
Works with Apache Kafka, Redpanda, MSK, and Confluent Cloud

Tables & functions it exposes

Once attached, the connector adds these objects to your SQL session.

Table kafka.topics: One row per topic with partition and offset metadata.
Table kafka.consumer_groups: Lag and assignment per consumer group.
Function read_topic(topic, from): Stream a topic as rows, decoded to typed columns.
Function produce(topic, rows): Write a result set back to a Kafka topic.
Secret kafka_credentials: SASL / TLS credentials stored as a DuckDB secret.

What teams build with Kafka

Ad-hoc debugging of what is actually flowing through a topic.

Join a live event stream against reference tables in your warehouse.

Backfill a table from a topic replay without standing up a pipeline.

Ready to attach Kafka?

Pricing

Attach & query

kafka.sql

-- Attach the Kafka worker from the Orchard
ATTACH 'kafka' AS kafka (
  TYPE vgi,
  LOCATION 'https://orchard.query.farm/kafka',
  TOKEN getenv('ORCHARD_TOKEN')
);

-- Tail the last 10 minutes of an orders topic
SELECT key, value:customer_id AS customer_id, value:total AS total, ts
FROM read_topic('orders', from => 'latest')
WHERE ts > now() - INTERVAL 10 MINUTE
ORDER BY ts DESC;

Requires the VGI extension. Set ORCHARD_TOKEN after subscribing.

Keep exploring