Connectors & destinations

A connector is a workspace-level link to an external system. It can act as a source — documents that arrive there are fed into a pipeline — and as a destination — the pipeline's results are written back to it. Connectors are managed in Settings → Integrations, on the Connectors tab, separately from any one pipeline, so one connector can serve many pipelines.

Google Drive, today ​

The only connector type today is Google Drive. A Google Drive connector is an authenticated Google account; it carries no folders of its own. Each pipeline chooses its folders from that account in the pipeline designer — an input folder on the Source stage and an output folder on the Destination stage.

A file dropped into a pipeline's input folder is processed by that pipeline and then moved to a processed/ subfolder, so the input folder doubles as a visible queue. A file rejected at intake — an unsupported type, for example — moves to failed/ with a .rejected.txt note beside it. When the job completes, the result is written to the pipeline's output folder as a JSON file, with an XLSX rendering of it alongside.

One connector, many pipelines ​

One connector can feed many pipelines, each watching a different folder in the same Drive account. The restriction is on the input folder: it can be the source of at most one published pipeline. Publishing a second pipeline that watches the same connector and folder is rejected with CONNECTOR_FOLDER_CONFLICT, and the Source stage's folder picker disables folders another published pipeline already watches. A pipeline's output folder must differ from its input folder — results written into the watched folder would be ingested again.

A connector cannot be deleted while any pipeline still uses it as a source or destination; the Connectors tab names the pipelines to disconnect first.

Web Response is always on ​

Independently of any connector, every pipeline ends in a Destination stage whose Web Response output is always on. The result of every job is retrievable over the Job Results API and visible in the API Playground — a connector adds a delivery target, it never replaces the API. A pipeline with no connector still returns its result; it simply has no external folder to write to.

The Destination stage ​

The Destination stage decides where a finished result goes. It carries the always-on Web Response and, optionally, one Deliver to Google Drive target: an active Google Drive connector plus an output folder. Choosing a different connector replaces the target; choosing none leaves Web Response only. The output format — for example JSON — is set on the Post Processing stage, not here; the Destination stage governs routing, not shape.

Delivery semantics ​

• On completion — a connector receives the result when the job finishes, not while it is still processing. A job held for review delivers once it completes after the review.
• One published pipeline per input folder — an input folder feeds a single published pipeline, which keeps a document's route unambiguous; the connector itself can serve many.
• Reconnect on lost access — if Google access is revoked or lost, the connector stops polling and delivering until it is reconnected from its card.

What exists today ​

Other connector types are not yet available. Until then, endpoints that need the result programmatically use the Web Response API, and outbound notifications use webhooks.