Parse Document - Unsiloed AI

Overview

The Parse Document endpoint processes PDF documents and breaks them into meaningful sections with detailed analysis including text extraction, image recognition, table parsing, and OCR data. This endpoint supports advanced customization options for fine-tuning the parsing behavior to match your specific use cases.

This endpoint returns a task ID for asynchronous processing. Use the GET /parse/ endpoint to check status and retrieve results when processing is complete.

Request

file

required

PDF file to process.

high_resolution

boolean

Whether to use high-resolution images for cropping and post-processing. (Latency penalty: ~7 seconds per page)

chunk_processing

object

Advanced chunking configuration options

segment_processing

object

Segment-specific processing configuration for different document elements

error_handling

string

Error handling strategy: “Continue”, “Fail” (default: “Continue”)

segmentation_strategy

string

Document segmentation strategy: “LayoutAnalysis” or “Page”

llm_processing

object

LLM processing configuration for enhanced content analysis

ocr_strategy

string

OCR processing strategy: “All”, “Auto”(default: “All”)

ocr_engine

string

OCR engine selection: “UnsiloedStorm”, “UnsiloedHawk” (default: “UnsiloedStorm”)

Advanced Configuration Options

Chunk Processing Configuration

The chunk_processing parameter allows fine-tuning of how document content is chunked:

{
  "chunk_processing": {
    "ignore_headers_and_footers": false,
    "target_length": 512,
    "tokenizer": {
      "Enum": "Word"
    }
  }
}

Parameters:

ignore_headers_and_footers (boolean, optional): Whether to exclude headers and footers from chunking.
target_length (number, optional): Target chunk length in tokens (default: 512)
tokenizer (object, optional): Tokenization strategy configuration with Enum values: “Word”, “Cl100kBase”, “XlmRobertaBase”, “BertBaseUncased”

Segment Processing Configuration

The segment_processing parameter allows customization of how different document segments are processed. Each segment type can be configured individually:

{
  "segment_processing": {
    "Table": {
      "html": "LLM",
      "markdown": "LLM",
      "extended_context": false
    },
    "Picture": {
      "crop_image": "All",
      "html": "LLM",
      "markdown": "LLM",
      "embed_sources": ["Markdown"]
    },
    "Formula": {
      "html": "LLM",
      "markdown": "LLM"
    },
    "Text": {
      "html": "LLM",
      "markdown": "LLM"
    },
    "SectionHeader": {
      "html": "LLM",
      "markdown": "LLM"
    },
    "Title": {
      "html": "LLM",
      "markdown": "LLM"
    },
    "Caption": null,
    "Footnote": null,
    "ListItem": null,
    "Page": null,
    "PageFooter": null,
    "PageHeader": null
  }
}

html (string, optional): HTML generation method: “LLM”, “Auto”
markdown (string, optional): Markdown generation method: “LLM”, “Auto”
extended_context (boolean, optional): Use the full page image as context for LLM generation
crop_image Controls whether to crop the file’s images to the segment’s bounding box. The cropped image will be stored in the segment’s image field. Use All to always crop, or Auto to only crop when needed for post-processing

Available Segment Types

Table: Tabular data processing
Picture: Image and graphic processing
Formula: Mathematical expression processing
Text: Regular text content processing
SectionHeader: Section heading processing
Title: Document title processing
Caption: Image/table caption processing
Footnote: Footnote processing
ListItem: List item processing
Page: Page-level processing
PageFooter: Footer processing
PageHeader: Header processing

LLM Processing Configuration

The llm_processing parameter configures language model processing:

{
  "llm_processing": {
    "fallback_strategy": "None",
    "max_completion_tokens": 1000,
    "model_id": "gpt-4",
    "temperature": 0.1
  }
}

Parameters:

fallback_strategy (string, optional): Fallback strategy when LLM processing fails: “None”,
max_completion_tokens (number, optional): Maximum tokens for LLM completion (default: 1000)
model_id (string, optional): LLM model identifier to use for processing
temperature (number, optional): Temperature setting for LLM processing (0.0 to 1.0)

Error Handling Strategies

error_handling (string, optional): How to handle processing errors:

“Continue”: Continue processing and report errors
“Fail”: Stop processing on first error

Segmentation Strategy Options

segmentation_strategy (string, optional): Document segmentation approach:

“LayoutAnalysis”: Analyzes pages for layout elements (e.g., Table, Picture, Formula, etc.) using bounding boxes. Provides fine-grained segmentation and better chunking.
“Page”: Treats each page as a single segment. Faster processing, but without layout element detection and only simple chunking.

OCR Strategy Options

ocr_strategy (string, optional): OCR processing strategy:

“All”: Process all text elements with OCR
“Auto”: Automatically determine when to use OCR

OCR Engine Options

ocr_engine (string, optional): Select the OCR engine for text recognition:

“UnsiloedStorm” (default): Fast processing, optimized for general documents
“UnsiloedHawk”: Higher accuracy, better for complex layouts and multilingual content

Engine Comparison:

UnsiloedStorm:
- Fast but noisy
- Faster processing time
- Good for standard documents
- Optimized for English text
- Lower resource usage
UnsiloedHawk:
- Slow but accurate
- Higher accuracy
- Better handling of complex layouts
- Superior multilingual support
- Longer processing time

Response

task_id

string

Unique identifier for the processing task

status

string

Initial task status (typically “Starting”)

file_name

string

Name of the uploaded file

created_at

string

Timestamp when the task was created

message

string

Status message about the task creation

configuration

object

Complete configuration used for processing

expires_at

string

Timestamp when the task expires (if expiration is set)

task_url

string

URL for accessing task status and results

Document Analysis Features

The parsing endpoint provides comprehensive document analysis including:

Text Extraction

Extracts text content with high accuracy, preserving formatting and structure.

Image Recognition

Identifies and analyzes images within documents, providing descriptions and metadata.

Table Parsing

Extracts tabular data with proper structure and formatting.

OCR Processing

Performs optical character recognition on text elements with confidence scores.

Section Detection

Automatically identifies different document sections like headers, body text, and captions.

Bounding Box Information

Provides precise coordinates for all extracted elements.

Advanced Content Processing

LLM-Enhanced Analysis: Uses language models for better content understanding
Multi-Format Output: Generates HTML, Markdown, and plain text versions
Context-Aware Processing: Maintains document context across segments
Intelligent Chunking: Creates semantically meaningful document chunks

curl -X 'POST' \
  'https://prod.visionapi.unsiloed.ai/parse' \
  -H 'accept: application/json' \
  -H 'api-key: your-api-key' \
  -H 'Content-Type: multipart/form-data' \
  -F 'file=@document.pdf;type=application/pdf' \
  -F 'high_resolution=true' \
  -F 'chunk_processing={"ignore_headers_and_footers": false, "target_length": 512, "tokenizer": {"Enum": "Word"}}' \
  -F 'segment_processing={"Table": {"html": "LLM", "markdown": "LLM", "extended_context": false}, "Picture": {"crop_image": "All", "html": "LLM", "markdown": "LLM", "embed_sources": ["Markdown"]}, "Formula": {"html": "LLM", "markdown": "LLM"}}' \
  -F 'error_handling=Continue' \
  -F 'segmentation_strategy=LayoutAnalysis' \
  -F 'ocr_strategy=All' \
  -F 'ocr_engine=UnsiloedStorm'

{
  "task_id": "e77a5c42-4dc1-44d0-a30e-ed191e8a8908",
  "status": "Starting",
  "file_name": "document.pdf",
  "created_at": "2025-07-18T10:42:10.545832520Z",
  "expires_at": "2025-07-18T11:42:10.545832520Z",
  "message": "Task created successfully. Use GET /parse/{task_id} to check status and retrieve results.",
  "task_url": "https://prod.visionapi.unsiloed.ai/parse/e77a5c42-4dc1-44d0-a30e-ed191e8a8908",
  "configuration": {
    "high_resolution": true,
    "chunk_processing": {
      "ignore_headers_and_footers": false,
      "target_length": 512,
      "tokenizer": {
        "Enum": "Word"
      }
    },
    "error_handling": "Continue",
    "segmentation_strategy": "LayoutAnalysis",
    "ocr_strategy": "All",
    "ocr_engine": "UnsiloedStorm"
  }
}

Retrieving Results

After the task is created, use the GET /parse/ endpoint to check status and retrieve results:

cURL

curl -X 'GET' \
  'https://prod.visionapi.unsiloed.ai/parse/{task_id}' \
  -H 'accept: application/json' \
  -H 'api-key: your-api-key'

Python

import requests
import time

def get_parse_results(task_id, api_key):
    """Monitor task and retrieve results when complete"""
    
    headers = {"api-key": api_key}
    status_url = f"https://prod.visionapi.unsiloed.ai/parse/{task_id}"
    
    # Poll for completion
    while True:
        response = requests.get(status_url, headers=headers)
        
        if response.status_code == 200:
            status_data = response.json()
            print(f"Task Status: {status_data['status']}")
            
            if status_data['status'] == 'Succeeded':
                return status_data  # Results are included in the same response
                    
            elif status_data['status'] == 'Failed':
                raise Exception(f"Task failed: {status_data.get('message', 'Unknown error')}")
                
        time.sleep(5)  # Check every 5 seconds

# Usage
task_id = "e77a5c42-4dc1-44d0-a30e-ed191e8a8908"
results = get_parse_results(task_id, "your-api-key")

Expected Results Structure

When the task completes successfully, the response contains comprehensive document analysis with enhanced processing:

{
  "task_id": "e77a5c42-4dc1-44d0-a30e-ed191e8a8908",
  "status": "Succeeded",
  "created_at": "2025-07-18T10:42:10.545832Z",
  "started_at": "2025-07-18T10:42:10.666328Z",
  "finished_at": "2025-07-18T10:42:23.560572Z",
  "message": "Task succeeded",
  "total_chunks": 2,
  "chunks": [
    {
      "segments": [
        {
          "segment_type": "Picture",
          "content": "eternal",
          "image": "https://s3.us-east-1.amazonaws.com/unsiloed-bucket/...",
          "page_number": 1,
          "segment_id": "1c60ecbd-b6da-493c-b3f6-6849337a981f",
          "confidence": 0.6062846,
          "page_width": 1191.0,
          "page_height": 1684.0,
          "html": "<p>The image displays a logo...</p>",
          "markdown": "Eternal Logo\n\nThe image displays a logo...",
          "bbox": {
            "left": 72.92226,
            "top": 62.030334,
            "width": 230.36308,
            "height": 55.395317
          },
          "ocr": [
            {
              "bbox": {
                "left": 63.753525,
                "top": 5.395447,
                "width": 164.45312,
                "height": 42.757812
              },
              "text": "eternal",
              "confidence": 0.9999992
            }
          ]
        },
        {
          "segment_type": "SectionHeader",
          "content": "Tax Invoice ORIGINAL For Recipient",
          "image": null,
          "page_number": 1,
          "segment_id": "9d60e48b-77ba-4a23-a0ac-95ee13c615ec",
          "confidence": 0.46558982,
          "page_width": 1191.0,
          "page_height": 1684.0,
          "html": "<h2>Tax Invoice ORIGINAL For Recipient</h2>",
          "markdown": "## Tax Invoice ORIGINAL For Recipient",
          "bbox": {
            "left": 486.9685,
            "top": 139.61847,
            "width": 241.29932,
            "height": 48.451706
          },
          "ocr": [
            {
              "bbox": {
                "left": 50.9729,
                "top": 3.4557495,
                "width": 46.046875,
                "height": 19.734375
              },
              "text": "Tax",
              "confidence": 0.99999654
            }
          ]
        }
      ]
    }
  ]
}

Segment Types

The parsing API identifies and processes different types of document segments with enhanced processing:

Picture

Images and graphics within the document, including logos, charts, and illustrations. Enhanced with LLM-based description generation.

SectionHeader

Document headers and titles that define section boundaries. Processed with semantic understanding.

Text

Regular text content including paragraphs, sentences, and individual text elements. Enhanced with context-aware processing.

Table

Tabular data with structured rows and columns. Enhanced with LLM-based formatting and extended context options.

Caption

Text captions associated with images or figures. Processed with relationship awareness.

Formula

Mathematical equations and expressions. Enhanced with specialized formula processing.

Title

Document titles and main headings. Processed with enhanced formatting.

Footnote

Document footnotes and references. Processed with context linking.

ListItem

Bulleted and numbered list items. Processed with structure preservation. Each segment includes detailed metadata such as confidence scores, bounding boxes, OCR data, and formatted output in both HTML and Markdown with LLM enhancement.

Configuration Best Practices

For High-Accuracy Processing

{
  "high_resolution": true,
  "chunk_processing": {
    "ignore_headers_and_footers": false,
    "target_length": 512
  },
  "segment_processing": {
    "Table": {"html": "LLM", "markdown": "LLM"},
    "Picture": {"crop_image": "All", "html": "LLM"},
    "Formula": {"html": "LLM", "markdown": "LLM"}
  },
  "segmentation_strategy": "LayoutAnalysis",
  "ocr_strategy": "All",
  "ocr_engine": "UnsiloedStorm"
}

OCR Engine Selection Guide

Choose the appropriate OCR engine based on your document characteristics: Use UnsiloedStorm when:

Processing standard business documents
Speed is prioritized over accuracy
Working with primarily English text
Processing large volumes of documents
Resource efficiency is important
You can tolerate some noise in the output for faster processing

Use UnsiloedHawk when:

High accuracy is critical
Working with multilingual documents
Quality over speed is preferred
You need clean, accurate text extraction

Example configurations:

// Fast processing for standard documents
{
  "ocr_engine": "UnsiloedStorm",
  "ocr_strategy": "All",
  "high_resolution": false
}

// High accuracy for complex documents
{
  "ocr_engine": "UnsiloedHawk", 
  "ocr_strategy": "All",
  "high_resolution": true
}

Error Handling

Common Error Scenarios

Invalid API Key: Authentication failed
File Too Large: File exceeds size limits
Invalid Configuration: Malformed processing parameters
Server Error: Internal processing error
Processing Timeout: Task took too long to complete

Authorizations

api-key

string

header

required

Body

multipart/form-data

file

required

PDF file to process

high_resolution

boolean

Whether to use high-resolution images for cropping and post-processing

segmentation_strategy

string

Document segmentation strategy: 'LayoutAnalysis' or 'Page'

ocr_engine

enum<string>

default:UnsiloedStorm

OCR engine selection: 'UnsiloedStorm', 'UnsiloedHawk' (default: 'UnsiloedStorm')

Available options:

UnsiloedStorm,

UnsiloedHawk

ocr_strategy

enum<string>

default:All

OCR processing strategy: 'All', 'Auto' (default: 'All')

Available options:

All,

Auto

Response

200 - application/json

Successful response

task_id

string

Unique identifier for the parsing task

status

string

Initial task status

Core Endpoints

Job Management

​Overview

​Request

​Advanced Configuration Options

​Chunk Processing Configuration

​Segment Processing Configuration

​Available Segment Types

​LLM Processing Configuration

​Error Handling Strategies

​Segmentation Strategy Options

​OCR Strategy Options

​OCR Engine Options

​Response

​Document Analysis Features

​Text Extraction

​Image Recognition

​Table Parsing

​OCR Processing

​Section Detection

​Bounding Box Information

​Advanced Content Processing

​Retrieving Results

​Expected Results Structure

​Segment Types

​Picture

​SectionHeader

​Text

​Table

​Caption

​Formula

​Title

​Footnote

​ListItem

​Configuration Best Practices

​For High-Accuracy Processing

​OCR Engine Selection Guide

​Error Handling

​Common Error Scenarios

Authorizations

Body

Response

Overview

Request

Advanced Configuration Options

Chunk Processing Configuration

Segment Processing Configuration

Available Segment Types

LLM Processing Configuration

Error Handling Strategies

Segmentation Strategy Options

OCR Strategy Options

OCR Engine Options

Response

Document Analysis Features

Text Extraction

Image Recognition

Table Parsing

OCR Processing

Section Detection

Bounding Box Information

Advanced Content Processing

Retrieving Results

Expected Results Structure

Segment Types

Picture

SectionHeader

Text

Table

Caption

Formula

Title

Footnote

ListItem

Configuration Best Practices

For High-Accuracy Processing

OCR Engine Selection Guide

Error Handling

Common Error Scenarios