Process documents with Layout Parser

Layout Parser extracts document content elements like text, tables, and lists, and creates context-aware chunks that facilitate information retrieval in generative AI and discovery applications.

Layout Parser features

Parse document layouts. You can input HTML or PDF files to Layout Parser to identify content elements like text blocks, tables, lists, and structural elements such as titles and headings. These elements help define the organization and hierarchy of a document with rich content and structural elements that can create more context for information retrieval and discovery.
Chunk documents. Layout Parser can break documents up into chunks that retain contextual information about the layout hierarchy of the original document. Answer-generating LLMs can use chunks to improve relevance and decrease computational load.

Taking a document's layout into account during chunking improves semantic coherence and reduces noise in the content when it's used for retrieval and LLM generation. All text in a chunk comes from the same layout entity, such as a heading, subheading, or list.

Limitations

The following limitations apply:

Online processing:
- Input file size maximum of 20 MB for all file types
- Maximum of 15 pages per PDF file
Batch processing:
- Maximum single file size of 40 MB for PDF files
- Maximum of 500 pages per PDF file

Layout detection per file type

The following table lists the elements that Layout Parser can detect per document file type.

File type	Detected elements	Limitations
HTML	paragraph, table, list, title, heading, page header, page footer	Be aware that parsing relies heavily on HTML tags, so CSS-based formatting might not be captured.
PDF	paragraph, table, title, heading, page header, page footer	Tables spanning multiple pages might be split in two tables.
DOCX (Preview)	paragraph, tables across multiple pages, list, title, heading elements	Nested tables are not supported.
PPTX (Preview)	paragraph, table, list, title, heading elements	For headings to be identified accurately, they should be marked as such within the PowerPoint file. Nested tables and hidden slides are not supported.
XLSX/XLSM (Preview)	tables within Excel spreadsheets, supporting `INT`, `FLOAT`, and `STRING` values	Multiple table detection is not supported. Hidden sheets, rows, or columns might also impact detection.

Before you begin

To turn on Layout Parser, follow these steps:

Create a Layout Parser by following the instructions in Creating and managing processors.

The processor type name is LAYOUT_PARSER_PROCESSOR.
Enable Layout Parser by following the instructions in Enable a processor.

Send an online process request with Layout Parser

Input documents to Layout Parser to parse and chunk.

Follow the instructions for batch processing requests in Send a processing request.

Configure fields in ProcessOptions.layoutConfig in ProcessDocumentRequest.

REST

Before using any of the request data, make the following replacements:

LOCATION: your processor's location, for example:
- us - United States
- eu - European Union
PROJECT_ID: Your Google Cloud project ID.
PROCESSOR_ID: the ID of your custom processor.
MIME_TYPE: Layout Parser supports application/pdf and text/html.
DOCUMENT: The content to be split into chunks. Layout Parser accepts raw PDF or HTML documents, or parsed documents that were output by the Layout Parser.
CHUNK_SIZE: Optional. The chunk size, in tokens, to use when splitting documents.
INCLUDE_ANCESTOR_HEADINGS: Optional. Boolean. Whether or not to include ancestor headings when splitting documents.

HTTP method and URL:

POST https://LOCATION-documentai.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/processors/PROCESSOR_ID:process

Request JSON body:

// Sample for inputting raw documents such as PDF or HTML
{
  "rawDocument": {
    "mimeType": "MIME_TYPE",
    "content": "DOCUMENT"
  },
  "processOptions": {
    "layoutConfig": {
      "chunkingConfig": {
        "chunkSize": "CHUNK_SIZE",
        "includeAncestorHeadings": "INCLUDE_ANCESTOR_HEADINGS",
      }
    }
  }
}

To send your request, choose one of these options:

curl

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://LOCATION-documentai.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/processors/PROCESSOR_ID:process"

PowerShell

Save the request body in a file named request.json, and execute the following command:

$headers = @{  }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://LOCATION-documentai.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/processors/PROCESSOR_ID:process" | Select-Object -Expand Content

The response includes the processed document with layout and chunking information as Document.documentLayout and Document.chunkedDocument.

Python

For more information, see the Document AI Python API reference documentation.

To authenticate to Document AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.


from typing import Optional, Sequence

from google.api_core.client_options import ClientOptions
from google.cloud import documentai

# TODO(developer): Uncomment these variables before running the sample.
# project_id = "YOUR_PROJECT_ID"
# location = "YOUR_PROCESSOR_LOCATION" # Format is "us" or "eu"
# processor_id = "YOUR_PROCESSOR_ID" # Create processor before running sample
# processor_version = "rc" # Refer to https://cloud.google.com/document-ai/docs/manage-processor-versions for more information
# file_path = "/path/to/local/pdf"
# mime_type = "application/pdf" # Refer to https://cloud.google.com/document-ai/docs/file-types for supported file types


def process_document_layout_sample(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
) -> documentai.Document:
    process_options = documentai.ProcessOptions(
        layout_config=documentai.ProcessOptions.LayoutConfig(
            chunking_config=documentai.ProcessOptions.LayoutConfig.ChunkingConfig(
                chunk_size=1000,
                include_ancestor_headings=True,
            )
        )
    )

    document = process_document(
        project_id,
        location,
        processor_id,
        processor_version,
        file_path,
        mime_type,
        process_options=process_options,
    )

    print("Document Layout Blocks")
    for block in document.document_layout.blocks:
        print(block)

    print("Document Chunks")
    for chunk in document.chunked_document.chunks:
        print(chunk)



def process_document(
    project_id: str,
    location: str,
    processor_id: str,
    processor_version: str,
    file_path: str,
    mime_type: str,
    process_options: Optional[documentai.ProcessOptions] = None,
) -> documentai.Document:
    # You must set the `api_endpoint` if you use a location other than "us".
    client = documentai.DocumentProcessorServiceClient(
        client_options=ClientOptions(
            api_endpoint=f"{location}-documentai.googleapis.com"
        )
    )

    # The full resource name of the processor version, e.g.:
    # `projects/{project_id}/locations/{location}/processors/{processor_id}/processorVersions/{processor_version_id}`
    # You must create a processor before running this sample.
    name = client.processor_version_path(
        project_id, location, processor_id, processor_version
    )

    # Read the file into memory
    with open(file_path, "rb") as image:
        image_content = image.read()

    # Configure the process request
    request = documentai.ProcessRequest(
        name=name,
        raw_document=documentai.RawDocument(content=image_content, mime_type=mime_type),
        # Only supported for Document OCR processor
        process_options=process_options,
    )

    result = client.process_document(request=request)

    # For a full list of `Document` object attributes, reference this page:
    # https://cloud.google.com/document-ai/docs/reference/rest/v1/Document
    return result.document

Batch process documents with Layout Parser

Use the following procedure to parse and chunk multiple documents in a single request.

Input documents to Layout Parser to parse and chunk.

Follow the instructions for batch processing requests in Send a processing request.

Configure fields in ProcessOptions.layoutConfig when making a batchProcess request.
Input
The following example JSON configures ProcessOptions.layoutConfig.
```
"processOptions": {
  "layoutConfig": {
    "chunkingConfig": {
      "chunkSize": "CHUNK_SIZE",
      "includeAncestorHeadings": "INCLUDE_ANCESTOR_HEADINGS_BOOLEAN"
    }
  }
}
```
Replace the following:
- CHUNK_SIZE: The maximum chunk size, in number of tokens, to use when splitting documents.
- INCLUDE_ANCESTOR_HEADINGS_BOOLEAN: Whether to include ancestor headings when splitting documents. Ancestor headings are the parents of subheadings in the original document. They can provide a chunk with additional context about its position in the original document. Up to two levels of headings can be included with a chunk.

What's next

Review the processors list.
Create a custom classifier.
Use Enterprise Document OCR to detect and extract text.
Review Send a batch process documents request to learn how to handle responses.

Custom extractor with generative AI

Pretrained overview