OCR for developers.
The fastest way to extract data from documents using AI.
$
bun i document-sdkbun i document-sdkcurl api.ocrbase.com/extractPaddleOCR
activeGLM-OCR
Custom model
Works with any model
Swap between PaddleOCR and GLM-OCR, or bring your own VLM with a single config change.
parse(file)
extract(file, schema)
batchParse(files)
batchExtract(files, schema)
4 functions0 configready
4 functions, that's it
parse, extract, batchParse, batchExtract. No bloat, no complexity.
output.json
{
"name": str
"date": date
"total": num
"items": [ ]
}Structured outputs
Extract data from documents directly into typed JSON with schema validation.