Document split

Overview

The document splitting feature in Extract API is designed for scenarios where multiple documents of the same type are combined into a single file (e.g., a single PDF file containing five different invoices). If you have a document with multiple documents of different types, you can use Document Classify to split the document into multiple documents.

How it works

The separation logic is primarily schema-driven. The model analyzes the document to identify how many times the structure defined in your schema appears.

For instance, consider a single file containing five invoices from the same company, each with a unique invoice number:

• Broad Schema: If your schema only defines generic fields like issuing_company, the model may interpret the entire file as a single document and return one extraction result.
• Specific Schema: If your schema defines specific identifiers like issuing_company and invoice_number, the model detects five distinct sets of information and splits the output into five separate extraction results.

Minimum Unit

The minimum unit for document splitting is a page. The model splits documents at page boundaries and cannot split content within a single page (e.g., two receipts scanned onto a single page cannot be separated).

Document split in Information Extract vs Document Classify

Document splitting is available in both Information Extract and Document Classify, but they operate on different principles and serve different use cases.

Usage

To enable this feature, simply use the split option in your request. Each responses of splitted schema will be returned as a separate item in the choices list.

Request For client setup, check simple extraction for more details.

extraction_response = client.chat.completions.create(
    model="information-extract",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {"url": f"data:application/octet-stream;base64,{base64_data}"}
                }
            ]
        }
    ],
    response_format={
        "type": "json_schema",
        "json_schema": {
            "name": "document_schema",
            "schema": {
                "type": "object",
                "properties": {
                    "bank_name": {
                        "type": "string",
                        "description": "The name of bank in bank statement"
                    }
                }
            }
        }
    },
    extra_body={ 
        "split": True
    } 
)
print(extraction_response)

Response

{
  "id": "iex-AQZoWf2p5j6TO-AE",
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "logprobs": null,
      "message": {
        "content": "{\"bank_name\":\"First National Bank\"}",
        "role": "assistant",
        "function_call": null,
        "tool_calls": null
      }
    },
    {
      "finish_reason": "stop",
      "index": 1,
      "logprobs": null,
      "message": {
        "content": "{\"bank_name\":\"Global Trust Bank\"}",
        "role": "assistant",
        "function_call": null,
        "tool_calls": null
      }
    },
    {
      "finish_reason": "stop",
      "index": 2,
      "logprobs": null,
      "message": {
        "content": "{\"bank_name\":\"Metro Financial Bank\"}",
        "role": "assistant",
        "function_call": null,
        "tool_calls": null
      }
    }
  ],
  "created": 1742838017,
  "model": "information-extract-260904",
  "object": null,
  "system_fingerprint": null,
  "usage": {
    "completion_tokens": 27,
    "prompt_tokens": 951,
    "total_tokens": 978,
    "completion_tokens_details": null,
    "prompt_tokens_details": null
  }
}