Configuration Options

When creating a new Extraction Agent, the schema is the most important part. However, there are a few other options that can significantly impact the extraction process.

Schema Alignment and Extraction Target

These options determine how your schema is applied to the document:

Extraction Target: Determines the scope and granularity of extraction. Available options:
- PER_DOC (default): Schema is applied to the entire document, returns a single JSON object
- PER_PAGE: Schema is applied to each page independently, returns an array of JSON objects (one per page)
- PER_TABLE_ROW: Schema is applied to each entity in an ordered list (table rows, bulleted lists, etc.), returns an array of JSON objects (one per entity)
See the Core Concepts page for detailed guidance on when to use each mode.

Important Settings

Model Settings

Extraction Mode: The mode of extraction to use. Can be either FAST, BALANCED, MULTIMODAL, or PREMIUM. The default is MULTIMODAL. FAST mode is suitable for simple documents with OCR but no AI-based parsing - it is the fastest mode. BALANCED mode provides a good balance between speed and accuracy for most documents. MULTIMODAL mode (default) is suitable for visually rich documents with a mix of text, simple tables, and images. PREMIUM mode should be used for extracting data from complex tables and information-dense documents - it provides the highest accuracy with advanced OCR, complex table/header and layout detection.
Parse Model: The model to use for document parsing. If not provided, uses the default for the extraction mode. Parse models are only configurable in MULTIMODAL and PREMIUM modes. See table below for available models per mode.
Extract Model: The model to use for data extraction. If not provided, uses the default for the extraction mode. Extract models are only configurable in PREMIUM mode. See table below for available models.

Available Models by Extraction Mode

Extraction Mode	Model Type	Models Available	Additional Credits	Description	Default
FAST	-	Not configurable
BALANCED	-	Not configurable
MULTIMODAL	Parse Model	`gemini-2.0-flash`	None	Fast processing on medium complexity documents	✓
		`gemini-2.5-flash-lite`	None	Fast and small model similar to Gemini 2.0 Flash
		`gemini-2.5-flash`	+15 cred. / page	Successor to Gemini 2.0 Flash with improved performance
		`openai-gpt-4-1`	+25 cred.	-
		`gemini-2.5-pro`	+55 cred.	Ideal for visually complex documents
	Extract Model	Not configurable
PREMIUM	Parse Model	`gemini-2.5-pro`	None	Ideal for visually complex documents	✓
		`openai-gpt-4-1`	None	-
		`anthropic-sonnet-4.5`	+45 cred. / page	High-performance parsing for complex layouts
	Extract Model	`openai-gpt-4-1`	None	-	✓
		`openai-gpt-5-mini`	None	Longer context retention and reasoning
		`openai-gpt-5`	+30 cred.	Longer context retention and reasoning (++)

System Prompt

System Prompt: Any additional system level instructions for the extraction agent. Note that you should use the schema descriptions to pass field-level instructions, few-shot examples, formatting instructions, etc.

Page Range and Context Window

Page Range: Specify which pages to extract from by providing comma-separated page numbers or ranges (1-based indexing). For example, use 1,3,5-7,9 to extract pages 1, 3, pages 5 through 7, and page 9. You can also use ranges like 1-3,8-10 to extract the first three pages and pages 8 through 10. Page numbers are 1-based, meaning the first page is page 1. This option is useful when you only need to extract data from specific sections of large documents.
Context Window: Number of pages to pass as context for long document extraction. This is useful when extracting from large documents where you need context from surrounding pages. Use the ExtractConfig.num_pages_context argument from the SDK to configure this option.

Metadata Extensions

For additional extraction features that provide enhanced metadata and insights, see the Metadata Extensions page which covers:

Citations: Source tracing for extracted fields
Reasoning: Explanations for extraction decisions
Confidence Scores: Quantitative confidence measures

These extensions return additional metadata in the extraction_metadata field but may impact processing time.

Other Advanced Options

These features are currently available under Advanced Settings in the UI.

Chunk Mode: Controls how the document is split for processing when dealing with large documents that exceed context limits. Choose between:
- PAGE: Each page becomes a separate chunk for processing (default)
- SECTION: Document is split into semantic sections based on content structure like headers and natural boundaries Use the ExtractConfig.chunk_mode argument from the SDK to configure this option.
High Resolution Mode: Enable high resolution processing for better accuracy on small text and fine details. This option improves OCR quality and uses higher resolution for document pages. Use the ExtractConfig.high_resolution_mode argument from the SDK to enable this feature. Warning: This can slow down extraction processing time, use only when needed.
Invalidate Cache: Purge cached results and rerun the full extraction. By default, LlamaExtract caches parsing results for 48 hours. Enable this option to bypass the cache and ensure fresh processing, which will incur fresh billing charges. Use the ExtractConfig.invalidate_cache argument from the SDK to enable cache invalidation.

Setting Configuration Options

You can configure these options when creating an extraction agent using either the REST API or Python SDK.

Python SDK

First, install the Python SDK:

pip install llama-cloud-services

Here’s how to set various configuration options:

from llama_cloud_services import LlamaExtract
from llama_cloud import ExtractConfig, ExtractMode, ChunkMode, ExtractTarget

# Initialize the client
extract = LlamaExtract(api_key="YOUR_API_KEY")

# Define your schema
schema = {
    "type": "object",
    "properties": {
        "company_name": {"type": "string", "description": "Name of the company"},
        "revenue": {"type": "number", "description": "Annual revenue in USD"}
    }
}

# Create extraction configuration
config = ExtractConfig(
    # Schema alignment
    extraction_target=ExtractTarget.PER_DOC,   # PER_DOC, PER_PAGE

    # Model settings
    extraction_mode=ExtractMode.MULTIMODAL,    # FAST, BALANCED, MULTIMODAL, PREMIUM
    parse_model=None,                          # Optional: override parse model
    extract_model=None,                        # Optional: override extract model

    # System prompt
    system_prompt="Focus on the most recent financial data",

    # Page range and context
    page_range="1-5,10-15",                    # Extract from specific pages
    num_pages_context=3,                       # Number of context pages for long docs

    # Metadata extensions (see Metadata Extensions page for details)
    cite_sources=True,                         # Enable citations
    use_reasoning=True,                        # Enable reasoning (not available in FAST mode)
    confidence_scores=True,                    # Enable confidence scores (MULTIMODAL/PREMIUM only)

    # Advanced options
    chunk_mode=ChunkMode.PAGE,                 # PAGE, SECTION
    high_resolution_mode=True,                 # Enable for better OCR
    invalidate_cache=False,                    # Set to True to bypass cache
)

# Create extraction agent with configuration
agent = extract.create_agent(
    name="Financial Data Extractor",
    data_schema=schema,
    config=config
)

# Extract from a document
result = agent.extract("your-document.pdf")
print(result.data)

REST API

You can also configure these options using the REST API when creating an extraction agent:

curl -X 'POST' \
  'https://api.cloud.llamaindex.ai/api/v1/extraction/extraction-agents' \
  -H 'accept: application/json' \
  -H "Authorization: Bearer $LLAMA_CLOUD_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "name": "Financial Data Extractor",
    "data_schema": {
      "type": "object",
      "properties": {
        "company_name": {"type": "string", "description": "Name of the company"},
        "revenue": {"type": "number", "description": "Annual revenue in USD"}
      }
    },
    "config": {
      "extraction_target": "PER_DOC",
      "extraction_mode": "MULTIMODAL",
      "parse_model": null,
      "extract_model": null,
      "system_prompt": "Focus on the most recent financial data",
      "page_range": "1-5,10-15",
      "num_pages_context": 3,
      "cite_sources": true,
      "use_reasoning": true,
      "confidence_scores": true,
      "chunk_mode": "PAGE",
      "high_resolution_mode": true,
      "invalidate_cache": false
    }
  }'

Configuration Reference

Option	Type	Default	Description
Schema Alignment
`extraction_target`	string	`"PER_DOC"`	Extraction scope: `PER_DOC`, `PER_PAGE`, `PER_TABLE_ROW`
Model Settings
`extraction_mode`	string	`"MULTIMODAL"`	Processing mode: `FAST`, `BALANCED`, `MULTIMODAL` (default), `PREMIUM`
`parse_model`	string	`null`	Parse model override (optional)
`extract_model`	string	`null`	Extract model override (optional)
System Prompt
`system_prompt`	string	`null`	Additional system-level instructions
Page Range and Context
`page_range`	string	`null`	Page numbers/ranges to extract (1-based, e.g., "1,3,5-7")
`num_pages_context`	integer	`null`	Number of pages to pass as context for long documents
Metadata Extensions
`cite_sources`	boolean	`false`	Enable source citations
`use_reasoning`	boolean	`false`	Enable extraction reasoning (not available in FAST mode)
`confidence_scores`	boolean	`false`	Enable confidence scores (MULTIMODAL/PREMIUM only)
Advanced Options
`chunk_mode`	string	`"PAGE"`	Document chunking strategy: `PAGE`, `SECTION`
`high_resolution_mode`	boolean	`false`	Enable high-resolution processing
`invalidate_cache`	boolean	`false`	Bypass cached results

Note: Some options like confidence_scores are only available in specific extraction modes. See the individual option descriptions above for details.