Open Source
The Document AI Framework
Parse, extract, and classify documents with any AI provider. Self-host or use our managed cloud.
npm i @doclo/sdkFeatures
Production-ready extraction in a few lines
Schema Extraction
Define your output shape with a typed schema. Get structured, validated data back with full TypeScript inference.
Doclo Cloud
Doclo SDK
const flow = createFlow()
.step('categorize', categorize({
provider,
categories: ['Passport', 'ID Card',
'Utility Bill', 'Bank Stmt']
}))
.conditional('extract', (ctx) => {
const schema = schemas[ctx.category];\n return extract({ provider, schema });
})
.build();
OpenAI
Anthropic
Google
Mistral
Datalab
Reducto
+100 other leading VLM models via OpenRouterIntegrations
Any model. Any provider. Zero lock-in.
Switch between AI providers and OCR engines with a single line of config. No rewrites, no migration headaches. Your pipeline stays the same.
Doclo SDK
Full control in your own environment
The same engine that powers Doclo Cloud. Self-host on your infrastructure, version everything in git, integrate into any stack.
View DocumentationTypeScript-First
Full type safety from schema to output. Your IDE knows exactly what you're getting back.
Provider Abstraction
Write once, run on any provider. No rewriting when you switch models.
Consensus Voting
Run N extractions, take the majority vote per field.
Automatic Resilience
Fallback chains, circuit breakers, exponential backoff.
OpenTelemetry Native
Every extraction traced. Export to your existing stack.
Citations & Traceability
Track exactly where each extracted value came from. Line-level source references with confidence scores.
Building Blocks
Composable nodes for any document workflow
Parse, extract, categorize, split, and more. Each node does one thing well. Chain them together for complex pipelines.
See the full node referenceParse
Extract text and structure from documents using OCR or Vision Language Models.
Extract
Pull structured data from documents according to your JSON Schema.
Categorize
Classify and route documents by type automatically.
Trigger
Launch sub-flows and reusable components.
Split
Identify and separate multiple documents within a single PDF.
Conditionals
Branch workflows with IF/THEN logic.
Doclo works with a broad spectrum of formats including PDFs, images, and Office documents. The exact formats supported depend on the models and providers you select for your pipeline.
Get Started
Ship document AI this week, not next quarter
npm i @doclo/sdk