Skip to content

Solutions / Document extraction

From document
to usable data.

Extract structured fields, trace them to the source, and check the result before it reaches your application.

Try the invoice demo

Sign in to run sample invoices when the demo is online.

Structured JSON extraction is one use case for the Qwen3.8 27B private business assistant. The example below is illustrative, not a measured result for this model. Evaluate accuracy on your own documents. Managed API input includes text, png/jpeg/webp images, and pdfs (scanned or digital, up to 10 pages).

Review image and PDF API examples and limits
Illustrative invoice → structured output
{
  "supplier": "Northstar Office Supply",
  "invoice_number": "NS-2048",
  "total": 482.17,
  "currency": "USD"
}

Fictional example. Run the live demo to see model output.

Structured fields

An open model fills an invoice schema from sample text.

Source evidence

Review where each extracted value appears in the original.

Checks in code

Validate dates, currency, and totals outside the model.

Build your own workflow

Start with a vLLM workspace.

Bring your model and extraction code. The demo’s validation pipeline is a separate example; a workspace gives you the model API to build on. You manage image and document processing in your own workspace; this is separate from managed API file support.

Launch vLLM