RPA.DocumentAI: extract structured data from documents

Currently supported engines:

• Google: google (requires rpaframework-google)
• Base64: base64ai
• Nanonets: nanonets

Tasks

Process a real world PDF invoice (or its PNG counterpart) with the following tasks:

• Document AI Google: using Google engine
• Document AI Base64: using Base64 engine
• Document AI Nanonets: using Nanonets engine
• Document AI All: using all the available engines
• Document AI Work Items: using custom configured engines for multiple files

Secrets

We recommend using Control Room's Vault for storing the API keys you'd normally need to configure before being able to operate these libraries which need to authenticate in their external services before being able to predict a document.

The expected structure can be observed within our vault.yaml file and looks like so:

out of which we can understand the following:

• Google uses a JSON (serialized dictionary) describing your service account private key under the serviceaccount field.
  • Init Document AI
• Base64 needs two comma-separated values under base64ai, which unfolds into the account e-mail address and an API key generated under it.
  • Set Authorization
• Nanonets only requires a valid API key under nanonets.
  • Set Authorization

Use Init Engine with any of the above for a unified & simplified experience.

⚠️ At all times keep these values private (securely stored in our Vault) and treat them like passwords.