Pipelines
A pipeline is a saved workflow: where documents come from, the steps to run, and where the results go. Open Processor → Pipelines to manage your pipelines and policies.
Create a workflow#
Choose New custom pipeline to build one yourself, or select Set up on one of the policies for a guided form.
- Name the pipeline and select an input source.
- Choose when it runs.
- Add tools in the order they should run and configure each step.
- Choose a destination, or return the modified document to the editor.
- Test the steps, then select Create pipeline or Create paused.
For a complete walkthrough, see Create your first pipeline.
Inputs and triggers#
| Input | Triggers |
|---|---|
| Editor | Every upload or Every export; see Editor |
| Folder | Manual, schedule, or folder watch |
| S3 / SFTP / FTP / SMB | Manual or schedule |
| Webhook | Signed HTTP delivery; see Sources |
Schedules use intervals in minutes, hours, or days. Folder watchers also check periodically for missed changes. Server triggers continue without a browser open; Editor workflows run during uploads or exports.
Steps and supporting files#
Each tool must accept the previous step's output format. A tool can appear more than once, with separate settings for each step.
To call another service, select Add a tool and look under Send to another system; see Integrations. Use Ingestion for knowledge search, Classification to label documents with AI, and Routing to choose destinations by document properties or classification.
Upload supporting files, such as stamp images or signing certificates, in the step settings. They are saved for future runs and shared within the team. Files must be non-empty and no larger than 50 MB. Referenced files cannot be deleted until removed from the pipeline.
Enforce as policy#
A pipeline with the Editor as its input runs automatically for everyone in the team when they upload or export a file. Enforce as policy decides what happens when it fails:
| When it fails | Ordinary pipeline | Enforced as policy |
|---|---|---|
| On upload | The file stays as it was uploaded, and the failure is reported. | The file is blocked. The person can retry or close it. |
| On export | The file is downloaded unprocessed, with a warning. | The download is cancelled. |
Use it when a rule must hold for every document, such as redacting personal data before anything leaves the editor, or converting to PDF/A before archiving.
Turn on Enforce as policy at the top of the pipeline builder, or in a policy's setup form. Enforced pipelines show as Policy in the pipelines list. Only people who can manage pipelines can change it.
Test and inspect output#
Test with a file runs the current steps, including unsaved edits, and returns outputs for download. It does not write to the configured storage destination. External integration calls still run.
After checking the output, run one controlled live input to confirm delivery to storage.
Save, pause, and edit#
Open a pipeline to edit it, then select Save changes. A policy that still fits its setup form opens in that form; select Customise there to switch to the full builder.
Pausing stops future automatic runs. The pause/enable control takes effect immediately, independently of unsaved step changes.
Destinations#
Folder and S3 destinations save processed files, using a new filename when one already exists. RAG database destinations receive prepared chunks from an Ingestion workflow.
Editor workflows can return the modified document to the editor. Database delivery and chunk exports need a separate destination. With routing, the default destination receives unmatched files.
