ドキュメントの抽出と検証

This section shows how to build a process that downloads an invoice, extracts data with UiPath Document Understanding™, routes a Document Validation task to a user, and completes on approval.

To build this process, you need to achieve the following objectives:

• Create an RPA workflow to download an invoice, extract fields, and generate validation artifacts.
• Create an Action App to validate the extracted content.
• Model a Maestro process that runs the RPA workflow, hands off to a user for validation, and finishes on approval.

前提条件

• You need access to Maestro, Orchestrator (shared folder and storage bucket), Studio Web, and Apps.
• You need a Document Understanding extractor/classifier configured for invoices.
• You need a sample invoice URL (publicly reachable or hosted in your environment).
• Create or open a Solution in Studio Web, and add a Maestro process. Read how to model a process.
• Ensure you can publish to your target Orchestrator folder and have a defined storage bucket.
• Ensure your Document Understanding model (extractor/classifier) is available to the RPA workflow.

Step 1 - Building and publishing the RPA workflow in Studio Web

To build and publish the RPA workflow in Studio Web, take the following steps:

1. Create a new RPA project (e.g., Document Extraction).
2. Add thwe following activities:
   • Download File from URL. Add the invoice URL as input.
   • Extract Document Data. Use your DU model and include classification if applicable.
   • Create Document Validation Artifacts. Configure the shared Orchestrator folder and the storage bucket.
3. Test the workflow locally in Studio Web.
4. Publish the workflow to Orchestrator.

Step 2 - Building and publishing the Action App in Studio Web

To build and publish the Action App in Studio Web, take the following steps:

1. Create an Action App task named Document Validation.
2. In the Action schema, add an input property (for example, validationData) of type ContentValidationData.
3. Add the validation control to the App, and set its data source to the validationData property.
4. Publish the Action App.

Step 3 - Modeling the agentic process in Maestro

To model the process in Maestro, take the following steps:

1. Add a Start event.
   • Use None (manual) for testing, or Message if you plan to trigger externally.
2. Add a Service task – Run extraction.
   • Implementation: Start and wait for RPA workflow.
   • Select the workflow you published in Step 1 - Build and publish the RPA workflow (Studio Web).
   • Map inputs (e.g., Invoice URL).
   • Capture outputs. Maestro exposes them as variables.
3. Add a User task – Validate Document Extraction.
   • Implementation: Create action app task.
   • Select the Document Validation Action App from Step 2.
   • Map the user task input to the extraction output (bind to the ContentValidationData payload).
4. Add an Exclusive gateway to branch on the user’s decision.
   • Approve: proceed to End.
   • Reject: route to a correction branch (optional) or loop back to RPA.

Step 4 - Publishing, running and validating the Maestro process

To publish, run and validate the Maestro process, take the following steps:

1. Publish and deploy the Maestro process. Read more about publishing.
2. Start a Debug run from the Start event.
3. When the Document Validation task is created, open it from the running instance.
4. Review the extracted fields in the validation control. Select Approve to continue the flow.
5. After completion, inspect Variables to see the stored documentValidation data and references to artifacts in your storage bucket.