Configuration - TextIn Intelligent Document Parsing

Use configuration options when the default output is not enough for your workflow. Enable options based on the document characteristics and the output your application requires.

Enabling every option does not necessarily improve the result. Some options increase response size or processing time.

Configuration object

{
  "document": {
    "password": "pdf-password"
  },
  "capabilities": {
    "include_hierarchy": true,
    "include_table_structure": true,
    "pages": false,
    "title_tree": false
  },
  "scope": {
    "page_range": "1-10"
  },
  "config": {
    "force_engine": "textin"
  }
}

Section	Purpose
`document`	Document-specific settings, such as a PDF password
`capabilities`	Controls the output fields and parsing capabilities returned by the API
`scope`	Limits parsing to selected pages
`config`	Advanced configuration

Common configurations

RAG and search
Complex tables
Scanned or photographed pages

Use this configuration when you need structure-aware chunks, headings, and table content.

{
  "capabilities": {
    "include_hierarchy": true,
    "title_tree": true
  }
}

This returns:

Heading and section relationships
A document title tree
Structured table data

Use this configuration when tables need to be processed programmatically.

{
  "capabilities": {
    "include_table_structure": true,
    "table_view": "html"
  }
}

This returns row, column, cell, span, and coordinate information for table elements when available.

Use preprocessing options only when the document needs them.

{
  "capabilities": {
    "crop_dewarp": true
  }
}

Enable:

crop_dewarp when pages are rotated, curved, cropped unevenly, or photographed at an angle.
remove_watermark only when visible watermarks interfere with recognition.
include_char_details only when character-level information or confidence scores are required.

Capabilities

`include_hierarchy`

Return parent-child relationships between elements.

Field	Type	Default
`include_hierarchy`	boolean	`true`

When enabled, elements may include relationship metadata such as parent and child element identifiers. Use for:

RAG chunking with context preservation
Document navigation
Hierarchical content analysis

`include_table_structure`

Return detailed table structure with rows, columns, cells, spans, and cell coordinates.

Field	Type	Default
`include_table_structure`	boolean	`false`

Use for:

Programmatic table processing
Financial report analysis
Database imports
Extracting specific cells

`include_inline_objects`

Return formulas, handwriting, checkboxes, and inline images within text when detected.

Field	Type	Default
`include_inline_objects`	boolean	`false`

Use for:

Scientific papers
Mathematical documents
Forms with checkboxes
Mixed-content documents

`include_image_data`

Return image URLs or image data for Image elements.

Field	Type	Default
`include_image_data`	boolean	`false`

Use for:

Downloading images from documents
Image processing pipelines
Visual content extraction

`include_char_details`

Return character-level coordinates and confidence scores when available.

Field	Type	Default
`include_char_details`	boolean	`false`

Use this only when you need character-level information. It can significantly increase response size.

`pages`

Return page-level metadata.

Field	Type	Default
`pages`	boolean	`false`

Page metadata can include page dimensions, rotation, and element references.

`title_tree`

Return a document outline based on detected headings.

Field	Type	Default
`title_tree`	boolean	`false`

Use for:

Table of contents generation
Section-aware navigation
Structure-aware chunking

`table_view`

Choose the text representation for table elements.

Field	Type	Values
`table_view`	string	`markdown`, `html`

Use html when complex tables cannot be represented clearly in Markdown.

`remove_watermark`

Attempt to remove visible watermarks before recognition.

Field	Type	Default
`remove_watermark`	boolean	`false`

Enable this only when watermarks interfere with text recognition.

`crop_dewarp`

Correct skewed, curved, or photographed pages before recognition.

Field	Type	Default
`crop_dewarp`	boolean	`false`

Enable this when pages are rotated, curved, cropped unevenly, or photographed at an angle.

Scope

`page_range`

Parse only selected pages.

{
  "scope": {
    "page_range": "1-5,10,15-20"
  }
}

Use page_range to:

Test parsing on a subset of pages
Reduce processing time
Split a document that exceeds page limits
Process only the pages your application needs

Document settings

`password`

Pass the password for a password-protected PDF.

{
  "document": {
    "password": "pdf-password"
  }
}

Advanced configuration

`force_engine`

Force a specific parsing engine when supported.

{
  "config": {
    "force_engine": "textin"
  }
}

Only use advanced engine settings when instructed by TextIn support or when your application requires a specific engine behavior.

`engine_params`

Pass engine-specific parameters.

{
  "config": {
    "engine_params": {
      "parse_mode": "auto",
      "formula_level": 0,
      "image_output_type": "url",
      "recognize_chemical": false
    }
  }
}

Parameter	Type	Values	Description
`parse_mode`	string	`auto`, `scan`, `parse`, `vlm`	How the PDF is parsed. `auto`: The engine selects the mode automatically. Works for the widest range of documents. `scan`: Treat every page as an image. `parse`: Extract text from digital PDFs only. Fastest option. `vlm`: Parse with a vision-language model.
`formula_level`	int	`0`, `1`	Formula recognition level: `0` for standard, `1` for enhanced.
`image_output_type`	string	`url`, `base64`	How images are returned, as a `url` or `base64` string.
`recognize_chemical`	boolean	`true`, `false`	Extract chemical molecular structures from the document. Takes effect only in `vlm` mode. When enabled, the results appear in each element’s `text` field.

Engine parameters may change across engines and versions. Use them only when you have a documented requirement.

Performance tips

Use the default configuration for the first integration.
Enable table structure only when you need cell-level table data.
Enable character details only when you need character-level coordinates or confidence scores.
Use page_range to test large documents before parsing the full file.
Use the asynchronous API for large documents, batch processing, or workflows that do not need the result in the original request.

Synchronous requests and asynchronous jobs both have a maximum processing time of 10 minutes. See Supported Files & Limits for current limits.

Parse Quickstart

Parse your first document.

Response Format

Understand the response structure.

Async Processing

Process large documents asynchronously.

Supported Files & Limits

Review supported formats and limits.

​Configuration object

​Common configurations

​Capabilities

​include_hierarchy

​include_table_structure

​include_inline_objects

​include_image_data

​include_char_details

​pages

​title_tree

​table_view

​remove_watermark

​crop_dewarp

​Scope

​page_range

​Document settings

​password

​Advanced configuration

​force_engine

​engine_params

​Performance tips

​Related resources

Parse Quickstart

Response Format

Async Processing

Supported Files & Limits

Configuration object

Common configurations

Capabilities

`include_hierarchy`

`include_table_structure`

`include_inline_objects`

`include_image_data`

`include_char_details`

`pages`

`title_tree`

`table_view`

`remove_watermark`

`crop_dewarp`

Scope

`page_range`

Document settings

`password`

Advanced configuration

`force_engine`

`engine_params`

Performance tips

Related resources