Document Intelligence Client

This Python package offers a robust client for seamless interaction with the Document Intelligence API. It enables you to efficiently retrieve, manage, and process document data within your workspace.

Installation

To install the package directly from GitHub using pip, run the following command:

pip install sema4ai-di-client

Usage

After installing the package, you can import it and start using the DocumentIntelligenceClient class.

Importing the Package

from sema4ai.di_client import DocumentIntelligenceClient

Getting a Document Work Item

To retrieve a document work item, make sure you've set the required environment variables before initializing the client. Specifically, ensure that DOCUMENT_INTELLIGENCE_SERVICE_URL and AGENTS_EVENTS_SERVICE_URL is set.

When running in Sema4.ai Control Room these are all handled by the platform.

• Environment Variables: Make sure the required environment variables (DOCUMENT_INTELLIGENCE_SERVICE_URL, AGENTS_EVENTS_SERVICE_URL) are set in your environment before running the code.
• Workspace ID: If you are developing on local, then make sure to set workspace_id in DocumentIntelligenceClient(workspace_id='<your_workspace_id>'), which is optional and deduced on non-local environments from the URL passed.
• Initialize the Client: The client will automatically read the environment variables and initialize the connection.
• Error Handling: The example includes error handling and ensures the client is properly closed at the end.

Here's an example:

from sema4ai.di_client import DocumentIntelligenceClient

# Ensure environment variables are set for the service URLs
# Example:
# export DOCUMENT_INTELLIGENCE_SERVICE_URL='https://api.yourdomain.com'
# export AGENTS_EVENTS_SERVICE_URL='https://agents.yourdomain.com'

# Initialize the client
client = DocumentIntelligenceClient()

# Specify the document ID you want to retrieve
document_id = 'your_document_id'

# Get the document work item
try:
    document_work_item = client.get_document_work_item(document_id)
    if document_work_item:
        print("Document Work Item:")
        print(document_work_item)
    else:
        print("No document work item found for the given document ID.")
except Exception as e:
    print(f"An error occurred: {e}")
finally:
    client.close()  # Make sure to close the client connection

Available Methods

The DocumentIntelligenceClient offers several methods to interact with the Document Intelligence API and manage work items. Below are examples of the available operations:

• Get Document Type: Retrieve details about a specific document type.

  document_type = client.get_document_type(document_type_name)
  

• Get Document Format: Fetch the format of a document based on its type and class.

  document_format = client.get_document_format(document_type_name, document_class_name)
  

• Store Extracted Content: Store extracted content after processing a document.

  client.store_extracted_content(extracted_content)
  

• Store Transformed Content: Save content that has been transformed by a process.

  client.store_transformed_content(transformed_content)
  

• Store Computed Content: Submit content that has been computed after analysis.

  client.store_computed_content(computed_content)
  

• Get Document Content: Retrieve document content in various states, such as raw, extracted, transformed, or computed.

  content = client.get_document_content(document_id, content_state)
  

• Remove Document Content: Delete content for a document in a specific state.

  client.remove_document_content(document_id, content_state)
  

• Complete Work Item Stage: Mark a work item’s current stage as complete and move to the next stage.

  response = client.work_items_complete_stage(
      work_item_id='your_work_item_id',
      status='SUCCESS',  # or 'FAILURE'
      status_reason='optional_reason',  # Optional
      log_details_path='optional_log_path'  # Optional
  )
  

Dependencies

The package requires the following dependencies:

• urllib3 >= 1.25.3, < 2.1.0
• python-dateutil
• pydantic >= 2
• typing-extensions >= 4.7.1

These should be installed automatically when you install the package via pip.

Example Code Snippet

Below is an example code snippet if you are testing on prod.

from sema4ai.di_client import DocumentIntelligenceClient

client = DocumentIntelligenceClient()

# Fetch and print the document type
try:
    document_type = client.get_document_type("CounterParty Reconciliation")
    print(document_type)
except Exception as e:
    print(f"An error occurred: {e}")
finally:
    client.close()  # Ensure proper resource cleanup

Below is an example code snippet if you are testing on local.

from sema4ai.di_client import DocumentIntelligenceClient
import os

# Set the required environment variables within the Python code
os.environ['DOCUMENT_INTELLIGENCE_SERVICE_URL'] = 'http://127.0.0.1:9080'
os.environ['AGENTS_EVENTS_SERVICE_URL'] = 'http://127.0.0.1:9080'

workspace_id = "<your Sema4.ai Control Room Workspace ID>"
client = DocumentIntelligenceClient(workspace_id=workspace_id)

# Fetch and print the document type
try:
    document_type = client.get_document_type("CounterParty Reconciliation")
    print(document_type)
except Exception as e:
    print(f"An error occurred: {e}")
finally:
    client.close()  # Ensure proper resource cleanup