Skip to main content
A connector is a source, a sink, or both. Sources read data from an upstream system, and sinks write it to a destination. Some connectors ship both as separate packages (connectors/postgres/source, connectors/postgres/sink). When embedding Filament, include a connector with a blank import:
Each package registers a fresh connector instance for every run. Its spec provides the name, configuration fields, capabilities, and maturity shown by the API and web app.

​
Maturity levels

Every connector is registered with one of three maturity levels.

​
Sources

The SaaS sources are not separate codebases. Each is a YAML manifest interpreted by one generic HTTP connector. Adding another SaaS API is usually a new manifest, not a new connector. The manifest format is documented in HTTP manifests.

​
Sinks

​
Where to go next

Sources

How each source reads, covering scan strategies, cursors, and CDC.

Sinks

How each sink writes, covering schemas, type maps, and atomicity.

Build your own

Implement the Source or Sink interface and register it.