@docdigitizer/langchain

LangChain document loader for the DocDigitizer document processing API.

Installation

npm install @docdigitizer/langchain @langchain/core

Usage

import { DocDigitizerLoader } from '@docdigitizer/langchain';

// Load a single PDF
const loader = new DocDigitizerLoader('invoice.pdf', { apiKey: 'dd_live_...' });
const docs = await loader.load();

console.log(docs[0].pageContent);       // JSON with extracted fields
console.log(docs[0].metadata);          // documentType, confidence, etc.

// Load all PDFs from a directory
const dirLoader = new DocDigitizerLoader('invoices/', { apiKey: 'dd_live_...' });
const allDocs = await dirLoader.load();

Configuration

const loader = new DocDigitizerLoader('invoice.pdf', {
  apiKey: 'dd_live_...',          // or set DOCDIGITIZER_API_KEY env var
  baseUrl: 'https://custom.api',  // optional
  timeout: 300000,                 // optional (ms)
  maxRetries: 3,                   // optional
  pipeline: 'CustomPipeline',     // optional
  contentFormat: 'json',           // "json" | "text" | "kv"
});

Document Metadata

Each LangChain Document includes metadata:

| Field | Type | Description | |-------|------|-------------| | source | string | File path of the processed PDF | | documentType | string | Detected document type (e.g., "Invoice") | | confidence | number | Classification confidence (0-1) | | countryCode | string | Detected country code (e.g., "PT") | | pages | number[] | Page numbers where document was found | | traceId | string | Unique trace identifier |

License

MIT