Build Guide (1h36min)

https://youtu.be/CL9B-xFTlR8

NOTE: This is a long build guide, but there are quick select chapters provided in the description of the the YouTube Video.

CHAPTERS 0:00 - Spec Review 1:48 - Step 1: Get topic from user 2:36 - Step 2: Let AI gather additional context for us. 7:20 - Step 3: Write search queries based on the previous info. 12:45 - Step 4: Search Google For Each Query & Scrape Each Result 38:15 - Step 5: Write the Report in JSON 57:18 - Step 6: Run a Workflow to Generate an images for each section of the report 1:08:59 - Step 7: Add Images to the Final Structured JSON 1:19:41 - Step 8: Generate an HTML Asset

AI Agent Spec:

This agent should be triggered directly on MindStudio.

Functionality:

• Asks the user what they’d like to research.
• Gather additional context to help refine the topic details. Hint: User Context Block
• Based on the topic provided, the refined topic details, an the current date, generate a JSON array of search queries. Make sure that the queries are diverse enough to get different results on each search.
• Iterate through the list of search queries. For each search query, run a sub-workflow that:
  • Searches Google
  • Iterate through google search results. For each result:
    • Scrape the URL
    • Return the Scraped Content and the URL
  • Return the scraped content and URL from ALL of the search results
• Using the the topic provided, the refined topic details, and the content from all of the search results, generate the research report in JSON format to be used in an HTML asset. The JSON should include:
  • A “title” for the Report (string)
  • A “subtitle” for the report (string)
  • An (object array) of “sections” that include:
    • An “h2” section header (string)
    • an (object array) of “subsections” that include:
      • An “h3” subsection header (string)
      • An (array) of paragraphs for the subsection
• Using the generated JSON. Run a sub-workflow to generate an image for each section. The sub-workflow should:
  • Use the content from the section to generate an image prompt.
  • Generate an image using the image prompt.
• Use an AI Model, or run a Custom Function to add the image url to the each section of the research report JSON.
• Generate an asset that uses the full research report JSON. The asset should look like an article page that includes:
  • The report title
  • The report subtitle
  • Multiple sections with
    • An h2 header
    • An image related to each section.
    • Subsections with
      • An h3 header related to the section header
      • paragraphs about the subsection
  • Bonus: At the bottom of the report it should list out all of the source URLs used in the research report.
• The final result should look something like this:
  • **👉 SAMPLE OUTPUT (this was all generated from a single AI Agent)**

<aside>

Submit Completed Agent:

Once you are finished, submit your Agent below 👇

Agent Build Submission Form

</aside>

Resources

🎬 Related Video Guides:

https://university.mindstudio.ai/1-core-building-principles/chaining-multiple-blocks-together