Intelligent Automation

Build Intelligent Document Processing Workflow

This lab guides you in creating an Intelligent Document Processing (IDP) system that automates document extraction and data processing using webhooks. You will integrate OCR, NLP, and real-time webhook triggers to enable seamless document handling and workflow automation.

Use case:

• Build Intelligent Document processing powered by Amazon Textract and Amazon Bedrock.

Pre-requisites

• S3 bucket with Purchase order sample.

Step 1 - Build Prompts in Prompt Playground.

1. Using Prompt Playground, create a PO (Purchase Order) extraction prompt to extract metadata from the purchase order using the prompt template “PO Extractor”

2. Rename and Save the prompt.

3. In the Test & Compare tab, choose the preferred models, assess their outputs, and identify the most optimal model. Subsequently, click "Select as Best Answer" to designate it as the Primary Model. Additionally, if necessary, you may assign an alternative model as the Fallback Model. However, selecting a Primary Model is mandatory.

4. Click "Save prompt run" button provided at the right corner.

5. Save and Publish the prompt.

6. Next create an agent prompt using prompt playground with the Webhook Notify Agent template.

   1. Refer "Webhook Notify Agent" prompt to enter Agent input.

   2. Save the prompt. After saving the prompt, the Tool tab becomes visible within the prompt.

   3. Add Message Tool with Email with the credentials.

7. In the Test & Compare tab, choose the preferred models, assess their outputs, and identify the most optimal model. Subsequently, click "Select as Best Answer" to designate it as the Primary Model. Additionally, if necessary, you may assign an alternative model as the Fallback Model. However, selecting a Primary Model is mandatory.

8. Click "Save prompt run" button provided at the right corner.

9. Save and publish the prompt

Step 2: Build Agent 2.0 Recipe.

1. Create a new Recipe with the type “Agent 2.0.”

2. Use the following nodes to create the recipe.

   1. Webhook (Copy Token in your notepad to be used later).

   2. Connector (Use Amazon S3).

   3. Processing (Select Amazon Textract with Extract Layouts,Extract Tables settings).

   4. Start

   5. Prompt (Select the Extract PO prompt created in Step 1).

      1. Enter the label as PO Extractor.

      2. Select the Document cache setting as “Ephemeral.”

      3. Select the “Messages” setting as “Last Message”.

   6. Agent(Select the Webhook Notify Agent).

      1. Enter the label as Webhook Notify Agent.

      2. Select the Document cache setting as “Ephemeral.”

      3. Select the “Messages” setting as “Last Message”

3. End

4. Save and Publish the recipe.

Step 3: Send an API request to initiate the workflow.

1. To initiate the processing of a purchase order

curl -X 'POST' \
  '<Webhook URL from the Webhook Tile>' \
  -H 'accept: application/json' \
  -H 'x-api-token: <YOUR API TOKEN copied in Webhook tile>' \
  -H 'Content-Type: application/json' \
  -d '{
    "files": [
        {
            "content_type": "application/pdf",
            "file_content": "",
            "file_path": "<S3 path>"
        }
    ],
    "input_message": "Extract PO",
    "metadata": {}
}'

1. Retrieve a webhook request by ID.

curl -X 'GET' \
  '</api/webhook/ request/ { request_id}>' \
  -H 'accept: application/json' \
  -H 'x-api-token: <YOUR API TOKEN copied in Webhook tile>'

3. Retrieve webhooks by recipe and filters.

curl -X 'GET' \
  '</api/webhook/recipe/{recipe_id}>' \
  -H 'accept: application/json' \
  -H 'x-api-token: <YOUR API TOKEN copied in Webhook tile>'

1. Go to https://app.karini.ai/recipe , Click “Actions” next to your recipe, and click “Webhook History.”

   1. Observe the Webhook history messages and observe Traces

Sample resources

Prompt:

1. PO Extractor

2. Webhook Notify Agent

Recipe: IDP

Refer following video to create recipe.