Choose the fields.
Start from a workflow or describe your own document. Edit field types, descriptions and required values.
Your documents.
Your data shape.
Turn the PDFs and images in your intake queue into the fields your software needs. Define a schema. Get structured JSON, confidence and validation.
25 free credits each month · No credit card
supplier-invoice.pdfSource document
NORTHLINE SUPPLY
Invoice · INV-00482 · Fictional demonstration document
Structured outputJSON
{
"supplier_name":"Northline Supply",
"invoice_number":"INV-00482",
"invoice_date":"2026-10-08",
"due_date":"2026-11-07",
"total_amount":2480.5
}
Define your fields. Real API results add model-reported confidence, validation and review status.
Select a source field to inspect its JSON value. Expected output shown; no live extraction.Explore this workflow
PDFs and imagesSchema-shaped JSONConfidence and validationBranded upload portals
Start with the work
on your desk.
Every document has a next step. Explore fields, sample documents and review checks built around yours, then make the schema your own.
Explore the workflow libraryLease and rent schedules
Lease dates, parties and stated rent periods for your register.
Funding applications
Business, contact and requested amount fields for your intake record.
Accounts payable invoices
Supplier details, totals and line items for AP review.
Certificates of insurance
Policy dates, coverage limits and named insured fields.
Business bank statements
Dates, balances and transaction rows for lending intake.
Bills of lading
Shipment references, cargo, origin and destination fields.
A document pipeline.
Without the project.
One shared extraction endpoint handles the document reading. Your schema defines the result; your application keeps the decisions.
Send the document.
Upload in the dashboard, use a branded portal, or call the API with a file and saved schema slug.
Inspect the result.
Read the data, confidence and validation. Review uncertain fields before passing them downstream.
curl -X POST https://dokyumi.com/api/v1/extract \
-H "Authorization: Bearer YOUR_API_KEY" \
-F "file=@document.pdf" \
-F "schema=YOUR_SAVED_SCHEMA_SLUG"A result you
can work with.
Useful data needs a clear contract. Inspect the same output in your API response and extraction dashboard.
Test your first documentdata · The fields you asked for
Return an object shaped by your schema, including the tables and line items you define.
confidence · Know where to look closer
Keep model-reported field scores attached to the data. Use them alongside your own review checks.
validation · Keep exceptions visible
Inspect schema errors and low-confidence fields. A processing status is distinct from a business approval.
Industry classification
at application time.
For lending teams that need SIC codes, the dedicated classifier reads applications and bank statements and returns a proposed code, evidence and a review signal.
Try the SIC classifierCapacity that grows
with your intake.
Start with 25 free credits each month. Starter includes 500 credits for $99/month; Growth includes 3,000 for $499/month.
Compare plans and limitsA few useful answers.
What does Dokyumi return?
Dokyumi returns structured JSON shaped by the schema you define, plus model-reported field confidence, validation details, and a completed or review status. Your software decides what to do with the reviewed data.
Can I try my own documents?
Yes. Create a free account, open a workflow’s starter schema, choose the fields you need, and send a supported file from your dashboard or the API. Account documents are retained according to the privacy policy. The examples on this page are synthetic.
Which file formats work?
PDF, JPEG, PNG, TIFF and WEBP files up to 20MB. Both digital PDFs and scanned documents are supported. Evaluate representative layouts and scan quality before automating a document class.
How do extraction credits work?
One credit covers one document of up to 5 pages. Longer documents use additional credits in five-page blocks, rounded up. The free plan includes 25 credits each month. Self-service documents can be up to 50 pages.
Do I need to build an OCR pipeline?
No. Define your fields and send the file and your saved schema slug to POST /api/v1/extract. Dokyumi handles document reading and schema-based extraction. You keep control of business rules, review and integrations.
What happens when I send the same document again?
Identical documents can reuse cached OCR to reduce document reading work. Every extraction still consumes credits, including repeat files.
Can customers upload documents without using my API key?
Yes. Create a branded upload portal linked to a schema. Customer submissions appear in your account and support webhook delivery. Every plan includes portals; limits depend on the plan.
Watch the document-to-JSON walkthrough
See the document-to-JSON workflow. Example data is illustrative.
Read the video transcript
Every new document layout can turn a simple data task into another parsing project. Your team needs usable fields, without rebuilding the extraction pipeline.
With Dokyumi, start with the data you actually need. Define a schema for vendor name, invoice number, total, and due date.
Send a supported PDF or image, your schema slug, and an API key to one shared extraction endpoint.
Get structured JSON shaped to those fields, plus a model-reported confidence map and validation details.
A response can be completed or marked for review. Inspect errors and uncertain fields before your software takes the next step.
Keep your own business checks and routing. Test real document layouts before connecting the output to an accounting system, workflow, or application.
Start with twenty-five free credits each month. One credit covers a document up to five pages. Longer documents use additional credits. No credit card required.
Create your free account, define your first schema, and test a document. Open the docs for the request and response contract. Build your workflow with Dokyumi.
Make the next document
the easy one.
Choose a workflow, edit its schema, and test a representative file.