📘 Creating a Custom Extractor

📑 What is an Extractor and Why Use It?

In the UpBrains AI platform, an Extractor is an AI-powered tool that allows you to automatically extract structured data from unstructured documents like:

• PDFs  (e.g., invoices, contracts, receipts, etc.),

• Images (PNG, JPG, etc.),  

• Text content (email body, text files, etc), and/or 

• The total email content including the email body and all its attachments.

✅ Key Use Cases:

• Automatically capture key information from documents and emails

• Provide clean, structured inputs to AI models or analytics dashboards

• Power automations based on document content

• Reduce manual data entry and errors
  
  

✳️ Step-by-Step: How to Create a Custom Extractor

🟩 Step 1: Choose an Extractor Type

In Admin Workspace, click on Extractors. You can either use a Prebuilt Extractor or create a Custom Extractor.

🔸 Option 1: Use a Prebuilt Extractor

UpBrains AI provides a set of prebuilt extractors for common use cases (e.g., invoice extraction, customer info, etc.). These extractors are already published and ready to be used inside your AI agents under the Extract Information action.

To use a prebuilt extractor:

1. From the left sidebar, go to Extractor.

2. Browse the list of Prebuilt Extractors.

3. Click on Test button to test it.

4. Upload a sample file or text and click Extract to test it.

5. If it works for your use case, you can directly add it to your agent's workflow—no action is needed.

   
   

   
   

   

🔸 Option 2: Create a Custom Extractor

If you need to extract specific fields unique to your business or documents, create a custom extractor.

To create a custom extractor:

1. From the left sidebar menu, click on Extractor.

2. Then click Create Custom Extractor.

3. You’ll be guided through a step-by-step process to define how information should be extracted from your files or text.

   The process of creating an extractor is AI-powered by itself. You can start by a prompt that describes your desire extractor, or Skip (using Skip button on the top-right corner of the screen) to go to the interactive mode for defining a custom extractor.

   
   

Describe your desired header and line item fields, if application, and Click Generate for AI to start creating your extractor and proceed to the next screen.

🟦 Step 2: Review Extractor Setup

You'll land on on a page where its shows the generated extractor specification including Extractor Setup, Header Fields and Line Item files.

Extractor Setup:

🔹 Extractor Name: Shows an AI generated name for your extractor. You can edit it as needed.

 🔹 Document Extraction Guide: This is a placeholder for providing specific instructions to AI and describe the document as a whole, by sharing domain information, if needed. Leave this part blank for common documents such as invoices, purchase orders, etc.

Click to collapse this section after reviewing it.

🟦 Step 3: Review Header Fields

The next section to review is "Header Fields". Here, you will see the list of key header information you would like to extract from the document or the email. Review them for accuracy.

Using Prompts to Update your Extractor

You can use prompts to request AI to add/update/remove fields from the header fields section. Be specific about the ask.

Using UI to Update your Extractor

You can also use the interactive mode, using the UI, to edit the generated extractor specification.

To add a key:

1. Click +Add Key

2. Fill in the following fields:

   • Key Name (e.g., Invoice Number)

   • Type (Text, Date, Number, Address, Boolean)

   • Multiple Values? (Indicate if this field may have multiple values in the document and you would like to extract all different values)

   • Description (Optional context for this field)

🔵 Click on checkmark icon to save your changes.

🟦 Step 4: Review Line Item Fields

This step is useful if your documents include lists or tables (e.g., line items in invoices). Example line item keys are Description, Unit Price, and/or Quantity in an invoice.

To add line item fields:

1. Click +New Line Item

2. Enter:

   • Key Name

   • Type

   • Description

🔵 Once done, click Save and Test

🟦 Step 5: Test & Publish

✅ Test your extractor by uploading a file or entering a sample text (such as the body of an email).

This allows you to verify that the extractor correctly identifies and pulls out the intended data fields.

Steps:

1. Upload a test file or input text
You can upload a PDF, image, or .txt file, or simply paste raw text (e.g., the content of an email).
The extractor will automatically process it and extract the values.

2. Inspect the extracted values
Review the output to see if the correct data has been captured.
If you notice any incorrect or missing information, go back and edit your extractor setup. You may need to adjust the field descriptions or data types.

3. Re-test if needed
If you made changes to the extractor, come back to this test page and click Extract again.
This will use the latest version of your extractor to process the file or text.

4. Publish your extractor
Once you're satisfied with the results, click Publish to finalize it.
Your extractor is now ready to be used in an AI Automation Agent (in the "Information Extraction" step).

📝 Note:

After publishing, each extractor will have two versions:

• Draft: A working version that you can continue to edit, test, and improve.

• Published: The locked version that is used in live automation by AI agents. You can now see your custom extractor in the list of Extractors in Information Extraction action in Agent Flow Builder.

✅ Step 6: Access Extractor in "My Extractors"

Once published, your extractor will appear in the My Extractors list. From there, you can:

• Edit or update it (the Draft version) and publish it when satisfied with the changes

• Connect it to automation flows (the published version)

• Use it via API — by calling the Live URL of the Webhook URL of your AI agent