Skip to main content

​
What is the Parser?

The parser extracts document structure (blocks, layout) before AI extraction. This structured representation helps the AI understand document context and relationships between elements. Parser Output Includes:
  • Text blocks with coordinates
  • Layout structure (headers, footers, tables)
  • Reading order
  • Document hierarchy

​
Parser Toggle

​
use_parser

Controls whether to parse document structure before extraction.
When to use:
  • Multi-page documents
  • Complex layouts (tables, forms)
  • Documents requiring high accuracy
  • When you need citations
  • When using advanced chunking strategies
Benefits:
  • Higher extraction accuracy
  • Better context understanding
  • Supports all chunking strategies
  • Enables citations
  • Handles complex layouts better
Trade-offs:
  • Slower processing
  • Higher cost per document
{
  "config": {
    "use_parser": true,
    "parser_mode": "PLUS"
  }
}

​
Parser Modes

When use_parser: true, you can select the parser quality level with parser_mode.

​
Available Modes

LITE

Fastest, lowest cost (Default)Basic parsing for simple documents
  • Citations automatically disabled
  • Good for receipts, simple forms

PLUS

BalancedGood accuracy for most documents
  • Handles tables and forms well
  • Best for general use

PRO

Highest accuracyBest quality for complex documents
  • Advanced layout analysis
  • Best for contracts, complex invoices

​
Mode Comparison

FeatureLITEPLUSPRO
SpeedFastestFastSlower
CostLowestMediumHighest
AccuracyGoodBetterBest
Citations❌ Disabled✅ Supported✅ Supported
TablesBasicGoodExcellent
Complex LayoutsBasicGoodExcellent
Best ForSimple docsGeneral useComplex docs

​
Selecting Parser Mode

{
  "config": {
    "use_parser": true,
    "parser_mode": "PLUS"  // LITE, PLUS, or PRO
  }
}
If not specified, parser_mode defaults to LITE.

​
Impact on Other Features

​
Citations

Citations require use_parser: true. Additionally, citations are automatically disabled when parser_mode: "LITE".
// ✅ Citations will work
{
  "use_parser": true,
  "parser_mode": "PLUS",  // or PRO
  "enable_citations": true
}

// ❌ Citations disabled (LITE mode)
{
  "use_parser": true,
  "parser_mode": "LITE",
  "enable_citations": true  // Ignored
}

// ❌ Citations not supported
{
  "use_parser": false,
  "enable_citations": true  // Ignored
}
For citation details, see the Citations Configuration.

​
Chunking Strategies

Parser mode affects available chunking strategies: With Parser (use_parser: true):
  • ✅ VARIABLE - Dynamic chunking by character size
  • ✅ SECTION - Chunk by document sections
  • ✅ PAGE - Chunk by pages
  • ✅ PAGE_SECTIONS - Hybrid page and section chunking
  • ✅ BLOCK - Chunk by layout blocks
Without Parser (use_parser: false):
  • ✅ PAGE - Only strategy available
For chunking details, see the Chunking Configuration.

​
Template Configuration

Set parser mode in templates: 
{
  "name": "Invoice Extractor",
  "schema": {...},
  "config": {
    "use_parser": true,
    "parser_mode": "PLUS"
  }
}

​
Request Level (Override Template)

{
  "files": "https://example.com/document.pdf",
  "template_id": "INVOICE",
  "config": {
    "parser_mode": "PRO"  // Override template setting
  }
}

​
Next Steps

Chunking Configuration

Learn about chunking strategies

Citations

Track extraction sources

Configuration Overview

See all configuration options