Extract data with Docs
Read complex documents and keep the evidence beside every extracted field.
Provide a document
Upload a document as multipart form data with the field named file. Use the returned file ID in later calls, or supply a public document URL.
Files can be up to 50 MB and are retained for 24 hours. Supply pages to read selected pages, for example 1-5 or 3,7,12-15.
Define your extraction
Use parse for document markdown. Use extract for structured fields, defined by a JSON Schema, a field list, or a prompt.
curl --request POST \
--url 'https://factget.dev/docs/v1/extract' \
--header "Authorization: Bearer $FACTGET_DOCS_KEY" \
--header 'Content-Type: application/json' \
--data '{
"fileId": "YOUR_FILE_ID",
"prompt": "Extract the company name, reporting period, and total revenue."
}'Replace IDs and sample values with your own. Examples are not executed in your browser.
Review field statuses
Each extracted value carries a page, supporting quote, and status. Handle found, not_found, ambiguous, invalid, and missing explicitly in downstream review.
Review flagged fields against their source before relying on the output. Verification helps surface uncertainty; it does not remove the need for judgment.