Integrations

Connect your current stack

Feldera ingests streaming and database data, then writes to model serving or feature stores. No ETL middleware required.

Sources

Input connectors

Feldera consumes streaming topics, database CDC streams, and object store prefixes, processing every input incrementally.

Apache Kafka

Stream

Read Kafka topics as JSON, Avro, or Protobuf. Works with Confluent Cloud and self-hosted brokers. Avro and Protobuf schema registry integration is available.

Redpanda

Stream

Native Redpanda connector via the Redpanda API. No Kafka emulation layer required. Supports Redpanda Cloud and on-premises deployments with SASL authentication.

Postgres CDC

Database CDC

Slot-based logical replication CDC for Postgres 13+. Tracks inserted, updated, and deleted rows. Works with Amazon RDS, Cloud SQL, and Aurora Postgres.

MySQL CDC

Database CDC

Binlog CDC for MySQL 8.x with row-based replication. GTID offset tracking enables reliable exactly-once consumption. Supports Amazon RDS MySQL.

Amazon S3

Object Store

Incrementally ingest S3 prefixes with partition-aware scanning. Supports Parquet and JSON. New objects under the configured prefix are picked up automatically.

HTTP Push

Webhook

Send individual events through a Feldera-hosted HTTP endpoint. Fits custom event producers and webhook integrations that do not use a message broker.

Sinks

Output connectors

Feldera sends incremental results to model serving, feature stores, and downstream systems.

Feature API

REST API

Fetch feature rows by key through a managed REST endpoint. It returns the latest computed value, with bearer token auth, configurable CORS, and sub-100ms p99 latency.

Kafka Output

Stream

Send incremental deltas to a Kafka topic. Consumers receive changed rows only, rather than a full snapshot after each update. Supports JSON and Avro output.

Postgres Write

Database

UPSERT incremental results into a Postgres table. Suits ML infrastructure that reads feature values from relational tables, with batched writes and configurable flush intervals.

Feast Online Store

Feature Store

Write feature values to Feast online stores, including Redis and DynamoDB. Compatible with Feast 0.34+. The Feldera pipeline materializes updates and keeps the store current without batch runs.

S3 Parquet Export

Object Store

Write time-partitioned Parquet snapshots to S3 on a configurable schedule. Supports training pipelines needing periodic full snapshots alongside real-time inference features.

HTTP Webhook

Webhook

Send incremental change events to a configurable HTTP endpoint, either per event or in batches. HMAC signature verification secures webhook consumption.

Missing something?

Custom connector needed?

Cloud Scale customers may request custom connector onboarding. The open-source, documented connector SDK lets you build your own. Reach out to discuss your needs.