Document Understanding Cloud APIs Example

This project demonstrates how to digitize, classify, validate, and extract documents using UiPath Document Understanding API's.

Official Documentation

UiPath Document Understanding offers standalone capabilities, allowing integration with external tools and systems through APIs. This release includes APIs for Discovery, Digitization, Classification, Extraction, and Validation. Please take a look at the Official Documentation.

Project Features

• Interactive Menu to select your Project, Classifier, and Extractor(s)
• Digitize, Classify, and Extract Documents
• Optional Human In The Loop (HITL)
• Digitization Caching (7 Days)
• Classification CSV results
• Extraction CSV results

Process Flowchart

Requirements

• Python 3.12+
• requests library
• python-dotenv library
• questionary library

Setup

1. Clone the repository to your local machine:

   git clone https://github.com/nvpnathan/DU-Cloud-APIs.git

2. Navigate to the project directory:

   cd DU-Cloud-APIs

3. Create a Python virtual environment:

    python3 -m venv venv
    source venv/bin/activate

4. Install the required dependencies:

   pip3 install -r requirements.txt

5. Set up your environment variables by copying .env.example to .env file in the root directory and provide the APP_ID and APP_SECRET from your Cloud Envirnment:

APP_ID=
APP_SECRET=
AUTH_URL=https://cloud.uipath.com/identity_/connect/token
BASE_URL=https://cloud.uipath.com/<Cloud Org>/<Cloud Tenant>/du_/api/framework/projects/

Usage

Processing Documents

1. Place the documents you want to process in the specified folder (example_documents by default).

2. Run the main script main.py to process the documents:

   python3 src/main.py

3. Select your Document Understanding Project, Classifier (optional if extracting one document type only), and Extractor(s) (optional if classifying only).

4. Monitor the console output for processing status and any errors.

5. Classification and Extraction results will be printed to the console and saved in CSV format in the output_results folder.

File Structure

The project structure is organized as follows:

DU-Cloud-APIs/
│
├── src/
│   ├── main.py          # Main entry point for the application
│   ├── auth.py          # Authentication module for obtaining bearer token
│   ├── digitize.py      # Digitize module for initiating document digitization
│   ├── classify.py      # Classify module for document classification
│   ├── extract.py       # Extract module for document extraction
│   ├── validate.py      # Validate module for document validation
│   ├── api_utils.py     # Validate module for document validation
│   ├── config.py        # Configuration module for project variables, sqlite db creation
│   └── write_results.py # Utility module for and writing classification and extraction results to sqlite
│
├── tests/
│   ├── test_main.py     # Test for the main application entry point
│   ├── test_digitize.py # Test for the document digitization module
│   ├── test_classify.py # Test for the document classification module
│   ├── test_extract.py  # Test for the document extraction module
│   └── test_validate.py # Test for the document validation module
│
├── .env.example         # Example environment variables file
├── requirements.txt     # Python modules configuration file
├── example_documents/   # Folder containing example documents
├── generative_prompts/  # Folder containing Extraction and Classification Prompt Templates
└── output_results/      # Folder containing the CSV's of the Document Extraction Results

SQLite Usage

This project uses SQLite to store and manage various document processing results. The database schema includes the following tables:

1. documents: Stores metadata and processing stages for each document.

   • document_id: Unique identifier for the document.
   • filename: Name of the document file.
   • stage: Current processing stage of the document.
   • timestamp: Timestamp of the last update.
   • document_type_id: Type of the document.
   • digitization_operation_id: Operation ID for digitization.
   • classification_operation_id: Operation ID for classification.
   • classification_validation_operation_id: Operation ID for classification validation.
   • extraction_operation_id: Operation ID for extraction.
   • extraction_validation_operation_id: Operation ID for extraction validation.
   • digitization_duration: Duration of the digitization process.
   • classification_duration: Duration of the classification process.
   • classification_validation_duration: Duration of the classification validation process.
   • extraction_duration: Duration of the extraction process.
   • extraction_validation_duration: Duration of the extraction validation process.
   • error_code: Error code if any error occurred.
   • error_message: Error message if any error occurred.

2. classification: Stores classification results for each document.

   • id: Auto-incremented primary key.
   • document_id: Unique identifier for the document.
   • filename: Name of the document file.
   • document_type_id: Type of the document.
   • classification_confidence: Confidence score of the classification.
   • start_page: Starting page of the classified section.
   • page_count: Number of pages in the classified section.
   • classifier_name: Name of the classifier used.
   • operation_id: Operation ID for the classification.

3. extraction: Stores extraction results for each document.

   • id: Auto-incremented primary key.
   • filename: Name of the document file.
   • document_id: Unique identifier for the document.
   • document_type_id: Type of the document.
   • field_id: Identifier for the field.
   • field: Name of the field.
   • is_missing: Boolean indicating if the field is missing.
   • field_value: Extracted value of the field.
   • field_unformatted_value: Unformatted extracted value of the field.
   • validated_field_value: Validated value of the field.
   • is_correct: Boolean indicating if the extracted value is correct.
   • confidence: Confidence score of the extraction.
   • ocr_confidence: OCR confidence score of the extraction.
   • operator_confirmed: Boolean indicating if the value was confirmed by an operator.
   • row_index: Row index for table fields.
   • column_index: Column index for table fields.
   • timestamp: Timestamp of the extraction.

These tables are created and managed in the ensure_database function in src/config.py.

TODO

☐ Write Tests for Discovery API

☐ Perform validation outside of the workflow (optional)

☐ Create CSV output files from sqlite results

☐ Add unique batch_id for each run

☑ Moved Async requests to api_utils.py

☑ Added sqlit tables (documents, classification, extraction) for all classification, extraction, and validation results

☑ Write initial tests for core