Chunking Configuration

What is Chunking?

Chunking splits large documents into smaller pieces for processing. This enables:

Memory-efficient processing of large documents
Better extraction accuracy by focusing on relevant sections
Parallel processing for faster results
Handling documents that exceed model context limits

When to Use Chunking

Use Chunking

Documents over 10 pages
Complex multi-section documents
Memory-intensive processing
Extracting data scattered across pages

Skip Chunking

Single-page documents
Simple 2-3 page documents
When data is on first page only
Speed is critical priority

Enabling Chunking

Set use_chunk: true and provide chunking configuration:

{
  "config": {
    "use_chunk": true,
    "chunking_config": {
      "strategy": "PAGE",
      "target_chunk_size": 2
    }
  }
}

Chunking Strategies

Available strategies depend on whether parser is enabled.

With Parser (`use_parser: true`)

All chunking strategies available:

PAGE
VARIABLE
SECTION
PAGE_SECTIONS
BLOCK

Chunk by PagesGroups document into page-based chunks.

{
  "strategy": "PAGE",
  "target_chunk_size": 2  // 2 pages per chunk
}

When to use:

Document data organized by pages
Simple page-by-page processing
Invoices with line items per page

target_chunk_size: Number of pages per chunk (default: 1)

Dynamic Character-based ChunkingCreates chunks based on target character count.

{
  "strategy": "VARIABLE",
  "target_chunk_size": 1000  // 1000 characters per chunk
}

When to use:

Uniform chunk sizes needed
Documents without clear sections
Text-heavy documents

target_chunk_size: Target characters per chunk (default: 1000)

Chunk by Document SectionsGroups content by section headers.

{
  "strategy": "SECTION",
  "target_chunk_size": 2  // 2 sections per chunk
}

When to use:

Well-structured documents with headers
Contracts with sections/clauses
Reports with chapters

target_chunk_size: Number of sections per chunk (default: 1)

Requires document to have identifiable section headers.

Hybrid Page and Section ChunkingCombines page boundaries with section detection.

{
  "strategy": "PAGE_SECTIONS",
  "target_chunk_size": 1
}

When to use:

Documents with sections spanning multiple pages
Complex structured documents
Maintaining both page and section context

target_chunk_size: Chunking granularity (default: 1)

Chunk by Layout BlocksGroups by layout-detected blocks (paragraphs, tables, etc.).

{
  "strategy": "BLOCK",
  "target_chunk_size": 5  // 5 blocks per chunk
}

When to use:

Documents with mixed content types
Forms with distinct blocks
Complex layouts with tables and images

target_chunk_size: Number of blocks per chunk (default: 1)

Without Parser (`use_parser: false`)

Only PAGE strategy is available when use_parser: false.

{
  "config": {
    "use_parser": false,
    "use_chunk": true,
    "chunking_config": {
      "strategy": "PAGE",
      "target_chunk_size": 1
    }
  }
}

target_chunk_size Parameter

The meaning of target_chunk_size varies by strategy:

Strategy	target_chunk_size Meaning	Default
PAGE	Number of pages per chunk	1
VARIABLE	Target characters per chunk	1000
SECTION	Number of sections per chunk	1
PAGE_SECTIONS	Chunking granularity	1
BLOCK	Number of layout blocks per chunk	1

Start with default values and adjust based on document complexity and extraction results.

Strategy Selection Guide

Multi-page Invoices

Recommended: PAGEProcess 2-3 pages at a time for invoices with line items across pages.

{
  "strategy": "PAGE",
  "target_chunk_size": 2
}

Long Contracts

Recommended: SECTIONChunk by contract sections/clauses for better context.

{
  "strategy": "SECTION",
  "target_chunk_size": 2
}

Reports and Documents with Chapters

Recommended: SECTION or PAGE_SECTIONSMaintain section context while respecting page boundaries.

{
  "strategy": "SECTION",
  "target_chunk_size": 1
}

Forms with Mixed Content

Recommended: BLOCKGroup related form blocks together.

{
  "strategy": "BLOCK",
  "target_chunk_size": 5
}

ID Cards, Receipts (No Chunking)

Recommended: Disable chunkingSingle-page simple documents don’t need chunking.

{
  "use_chunk": false
}

Complete Configuration Examples

Template with Chunking

{
  "name": "Contract Extractor",
  "schema": {...},
  "config": {
    "use_parser": true,
    "parser_mode": "PRO",
    "use_chunk": true,
    "chunking_config": {
      "strategy": "SECTION",
      "target_chunk_size": 2
    }
  }
}

Request with Chunking Override

{
  "files": "https://example.com/long-contract.pdf",
  "template_id": "CONTRACT",
  "config": {
    "use_chunk": true,
    "chunking_config": {
      "strategy": "PAGE",
      "target_chunk_size": 3
    }
  }
}

Chunking without Parser

{
  "files": "https://example.com/document.pdf",
  "schema": {...},
  "config": {
    "use_parser": false,
    "use_chunk": true,
    "chunking_config": {
      "strategy": "PAGE",  // Only PAGE available
      "target_chunk_size": 1
    }
  }
}

Best Practices

Start Simple

Begin with PAGE strategy, adjust if needed

Match Document Structure

Choose strategy that aligns with document organization

Test Chunk Sizes

Experiment with target_chunk_size for optimal results

Monitor Results

Check extraction quality and adjust strategy

Very small chunk sizes may reduce accuracy by splitting related content. Very large chunks may exceed context limits.

Performance Considerations

Strategy	Processing Speed	Memory Usage	Accuracy
PAGE	Fast	Low	Good
VARIABLE	Fast	Low	Good
SECTION	Medium	Medium	Better
PAGE_SECTIONS	Medium	Medium	Better
BLOCK	Slower	Higher	Best

Next Steps

Parser Configuration

Understand parser requirements

Citations

Track extraction sources

Configuration Overview

See all configuration options

Introduction

Core Concepts

Configuration

API Reference

System

File Management

Document Processing

Job Management

Template Management

Account Usage

Fraud Detection

What is Chunking?

When to Use Chunking

Use Chunking

Skip Chunking

Enabling Chunking

Chunking Strategies

With Parser (`use_parser: true`)

Without Parser (`use_parser: false`)

target_chunk_size Parameter

Strategy Selection Guide

Complete Configuration Examples

Template with Chunking

Request with Chunking Override

Chunking without Parser

Best Practices

Start Simple

Match Document Structure

Test Chunk Sizes

Monitor Results

Performance Considerations

Next Steps

Parser Configuration

Citations

Configuration Overview

Introduction

Core Concepts

Configuration

API Reference

System

File Management

Document Processing

Job Management

Template Management

Account Usage

Fraud Detection

​What is Chunking?

​When to Use Chunking

Use Chunking

Skip Chunking

​Enabling Chunking

​Chunking Strategies

​With Parser (use_parser: true)

​Without Parser (use_parser: false)

​target_chunk_size Parameter

​Strategy Selection Guide

​Complete Configuration Examples

​Template with Chunking

​Request with Chunking Override

​Chunking without Parser

​Best Practices

Start Simple

Match Document Structure

Test Chunk Sizes

Monitor Results

​Performance Considerations

​Next Steps

Parser Configuration

Citations

Configuration Overview

What is Chunking?

When to Use Chunking

Enabling Chunking

Chunking Strategies

With Parser (`use_parser: true`)

Without Parser (`use_parser: false`)

target_chunk_size Parameter

Strategy Selection Guide

Complete Configuration Examples

Template with Chunking

Request with Chunking Override

Chunking without Parser

Best Practices

Performance Considerations

Next Steps