IDP
Templates that take a document, identify it, extract, process and validate its data in a pipeline and return a JSON, ready to test, version and call through the API.
IDP
IDP (intelligent document processing) turns documents — invoices, forms, receipts, PDFs and images — into structured data. Each use case is a template: a pipeline that takes a document, runs it through configurable stages and returns a JSON to the caller.
Use IDP templates when a process depends on reading documents that come from outside — from customers, suppliers, e-mails or portals — and you want that reading to be standardized, tested before it reaches production, and callable from systems, robots and agents.
The pipeline
Every document goes through the same stages, top to bottom: from the file received to the JSON returned.
| Stage | What it does |
|---|---|
| Document | Holds the template's reference document and sets which pages of each incoming document go through the pipeline. |
| Pre-processing | Image filters applied to the document before extraction, such as grayscale and sharpen. |
| Identification | Checks that the incoming document is the expected kind before anything is extracted, and can reject it. |
| Extraction | Gets the content out of the document with a single method: Roberty Cloud OCR or LLM OCR. |
| Post-processing | LLM or code steps that turn the extraction result into the JSON you want. |
| Validations | Checks the final result has to pass: code, a call to an API, or a robot. |
| Result | The final JSON, returned to whoever called the API and sent to the webhooks. |
Stages with nothing configured are simply skipped: with no filters, the document goes to extraction as it was sent; with no post-processing, the extraction result is returned as is; with no validations, every result is accepted. See Builder.
Draft and active version
Everything you edit in the template goes into the draft, which is saved automatically. The draft is what the Playground, Simulate API and the evaluation sets run, so you can test changes without affecting whoever already uses the template.
Publishing freezes the whole pipeline configuration into a version. The version marked as active is the one that answers in production: the API, the robot action and the agent action always run the active version. Until a version is published, the template does not process documents through the API. See Versions.
Environments
Every template run is recorded in one of three environments:
| Environment | What feeds it |
|---|---|
| Production | Calls through the API and webhook, the robot action and the AI agent action. |
| Development | Playground and Simulate API runs. |
| Testing | Evaluation set runs. |
See Runs.
Where templates live
IDP templates are created and organized in the workspace's Files tab, alongside robots, agents and code projects. An open template is laid out in tabs: Builder, Evaluation sets, Runs, Versions and Settings. At the top sit the active version (or Draft, if there is none yet), the Test menu — Playground, Simulate API and Test set — and the Publish button.
Only workspace developers and administrators can create and edit templates. See Roles and permissions.
A template can be turned on or off in Settings › General.
Some stages depend on features enabled on the workspace. Identification, LLM OCR and LLM post-processing steps need the generative AI feature; Roberty Cloud OCR needs the feature of the same name. If an option shows as locked, contact your account manager to have the feature turned on.