Introducing s-1: Sievly’s new document reading model
Extract
Pull the fields you need, with a source for every value
Define a schema and get back structured JSON from any document. Every value links to the page and region it came from.
POST /extract
Helping everyone from early-stage startups to global enterprises unlock their data.
Quarryline
Brindlewood
nimbleton
brightwick
vantrel
PLINTH & CO
OAKMERE
ARROWFEN
Halvory
pebblegate
Mirrowind
quillstack
Extract
Structured data you can verify
DEFINITION
Extract fills a schema you define from any document and returns each value with a citation and a confidence score.
WHO IT’S FOR
Teams automating intake, review and reporting in finance, insurance, legal and healthcare operations.
THE PROBLEM IT SOLVES
Prompting a general model for fields gives answers you cannot check. Extract shows exactly where each value came from, so reviewers can trust it.
Extract in the platform
Using Extract in your pipeline
Pair Extract with Parse or Split, or call it on its own.
#1
When you know the fields
Describe what you need once and run it on every document that arrives.
#2
When formats vary
The same schema works across vendors, templates and scans without per-format rules.
#3
When people review results
Citations and confidence let reviewers check only what needs checking.
Where teams use Extract
Schemas that replace manual data entry
Common jobs where Extract takes over the typing.
Invoices
Vendor, totals and line items from any invoice layout.
Financial statements
Balance sheet and income statement values with their periods.
Contracts
Parties, dates, governing law and key clauses.
Clinical forms
Patient details and answers, including handwriting.
Insurance documents
Policy numbers, coverages and limits from schedules.
Shipping documents
Shipper, consignee, weights and references.
Try a schema on your own documents in Workbench.
Why Extract
Why teams use Extract
01
Schemas, not prompts
Define fields and types once, then reuse them everywhere.
02
Citations on every value
Click through to the exact page and region.
03
Confidence you can route on
Send uncertain values to review automatically.
04
Built for long files
Find values that live pages apart in long documents.
How Extract works
How Extract works in four steps
■ STEP 01
Define a schema
List the fields and types you need.
schema.json
■ STEP 02
Send a document
Call Extract with the file and schema.
client.extract(file)
■ STEP 03
Review in Workbench
Check low-confidence values beside the page.
review queue
■ STEP 04
Ship the pipeline
Save it and call it by pipeline ID.
pipeline_id
Built for production
Built for production workloads
Extract runs on the same infrastructure as every Sievly endpoint, with the controls enterprise teams expect.
Built for regulated teams
Security controls cover the whole platform, with agreements available on Growth and Enterprise.
Deploy in your environment
Run Sievly in your own cloud when data cannot leave your network.
High-availability uptime
Built for steady throughput at any volume, with SLAs on Enterprise.
The rest of the platform
The endpoints around Extract
/parse
Parse
Turn any document into structured, cited content.
Read more →
/split
Split
Break long, multi-document files into logical parts.
Read more →
/classify
Classify
Label and route documents by type before they move on.
Read more →
/edit
Edit
Fill forms and update documents from plain language.
Read more →
/workbench
Workbench
Design, test and deploy pipelines without code.
Read more →
Further reading