Introduction
IDPForge turns unstructured documents into structured, validated data — through pipelines you configure, review queues your team works, and an API you call from anywhere.
What IDPForge does
You point IDPForge at documents — uploaded in Studio, posted to the API, or dropped in a connected Google Drive folder. It reads each file, can split a package into individual documents and classify each one, extracts and checks the fields your schema defines, and returns the result over the API or delivers it to a connected destination. When human review is turned on for a stage, anything the model is not confident about becomes an exception that a person clears in a review queue.
Pick your path
Production and test jobs
Every job is tagged production or test, and the Jobs page keeps the two on separate tabs so trial runs never mix with real traffic. Runs from a pipeline's Test tab and from the API Playground are test jobs. A job submitted through the API is production unless the request sets environment=test.
Test jobs use the same pipelines, schemas and webhooks as production jobs, and they consume credits the same way.
Every new workspace receives a one-time grant of free credits. Jobs run against that grant until it is spent — there is no daily job cap.
Accounts and billing
Work is metered in credits, not pages. Every job reserves credits up front, settles to what it actually used — never more than it reserved — and returns the rest to your balance. A job that fails releases its whole reservation, so it costs nothing; a cancelled job, or one that completes with errors, pays for the work it did. Inference runs on IDPForge-managed provider accounts, or on your own provider key.
Conventions in these docs
- How-to pages walk one task in the product UI, screen by screen.
- Reference pages describe what a thing is and every option it takes.
- Scenario pages follow one realistic end-to-end job from intake to delivery.
- Values you replace are written like
{job_id}. Anything in a code block is copy-safe.