Tutorial 3: Turning Website Content into an FAQ

​

Goal

1. Opening the target website(s).
2. Allowing the AI to explore and understand the content of these pages.
3. Instructing the AI within the Extract Data block to generate questions and answers based on the gathered information.

​

Key Blocks Used

• Open Websites
• Explore Content (optional, but helpful for broader site analysis)
• Extract Data (with generative instructions)

​

What You’ll Learn

• How to instruct the AI to explore a website to gather comprehensive information.
• How to write prompts for the Extract Data block to generate new content (FAQs) instead of just extracting existing text.
• A practical example of leveraging generative AI within a Jsonify workflow.

​

Steps

​

1. Create a New Workflow

• Navigate to your Jsonify Dashboard.
• Click the button sequence to create a new workflow: Create an empty workflow ➙ Extract.
  • This template will include Open Websites and Extract Data. We may add Explore Content.

​

2. Configure Open Websites Block

• Select the Open Websites block.
• Add the URL(s) of the website(s) you want to turn into an FAQ. For this example, let’s imagine we want to create an FAQ for Jsonify’s own website (or a specific section of it).
  • Example URL: https://jsonify.com/features (or any content-rich page)
• If adding multiple starting pages, list them here.
• Ensure the toggle switch next to each URL is ON.

Screenshot: Open Websites block configured with a target URL for FAQ generation

​

3. Add and Configure Explore Content Block

• If you want the AI to gather information from more than just the initial page(s) listed in Open Websites, add an Explore Content block. This helps the AI get a broader understanding.
• To add the block:
  1. Look at the block categories at the top of the editor.
  2. Select the Transform category.
  3. From the dropdown, click View More and select Explore Content.
  4. Click the + icon on the canvas after Open Websites.
• Select the Explore Content block to configure it:
  • Describe an action…: Explore the current site to understand its main features and offerings. Visit pages that seem to provide detailed information. Current page as well
  • How many pages to open?: The maximum number of pages you can specify is 50. However, for most cases, 5 to 10 pages will be sufficient..
  • Maximum number of steps: The same applies to the nuber of steps: the maximum is 50, but in most cases, you can specify 1 or 2.

Screenshot: Explore Content block configured for site exploration

​

4. Configure Extract Data Block (The Generative Step)

• This is where the FAQ generation happens. The Extract Data block will process the content from the page(s) visited (either directly from Open Websites or after Explore Content).
• Select the Extract Data block.
• What information should we scrape from each URL?: Select A list of items, as we expect to generate multiple Q&A pairs.
• Define the attributes for each FAQ item:

  NAME	EXAMPLE VALUE OR A LONGER DESCRIPTION
  question	A frequently asked question that can be answered from the page content.
  answer	A clear and concise answer to the corresponding question, derived from the page content.

• Add additional instructions and guidance…: This is the crucial part for generation.
  • Based on the content of the current page(s), generate a list of 3-5 frequently asked questions (FAQs). For each question, provide a concise answer using only information found on the page(s). Each question should end with a question mark. Format each as a distinct question and answer pair.
  • You can be more specific, e.g., Focus on questions related to pricing, features, and how to get started.

Screenshot: Extract Data block configured for FAQ generation

​

5. Run Your Workflow

• Click the Run manually button.
• The agent will:
  1. Open the initial URL(s).
  2. (If used) Explore the site to gather more context.
  3. For the content of the visited page(s), the Extract Data block will use its generative capabilities to create question and answer pairs.

​

6. Check the Results

• You’ll be redirected to the current run’s page.
• Once completed, click on “View as sheet”.
• You should see a table where each row is a generated question-answer pair.