Skip to main content

​
Initial Prompt

Give step by step instructions
If you know where your information is, tell Structify!e.g.: ‘Navigate to https://www.ycombinator.com/companies and extract the companies there’
See Structify Calls for info on how best to use our extraction toolkit
Once you have data, you can use Structify to enrich it with information from the web
The agent displays its thinking!
To unlock even more powerful uses, watch the agent’s thinking! This can give you insight on where it might go wrong, and allow you to prompt it in the correct direction!

​
Putting it all together

I want you to create a pipeline that extracts company information using ‘scrape_relationships’ from https://www.ycombinator.com/companies. I’m interested in company name, website, description and linkedin url. Once you’ve pulled in the companies, use enhance_columns starting at their website to gather information about each company’s pricing and product tiers.

​
Working with an Existing Pipeline

If you want or need a specific package installed, just ask! If you don’t know what this means, that’s okay, the ‘Fix with AI’ button will help
If you want to edit logic within a specific building block on the graph, mention in by name and tell us to edit that block specifically.
Any building block you edit, will cause it and all blocks below it to run again!
Most of the time, you’ll want to add new building blocks for new logic! For instance, if your initial prompt pulled in a bunch of databases and you want to aggregate information, you might ask ‘Write a new transform block to aggregate by X and Y, by Z. Order the results by date in ascending order’

​
Planning Mode

For more complex workflows, or business logic heavy ones, it can be good to give your prompt and ask the agent to ‘Plan’. This will have it draft up an implementation plan for you, ask clarifying questions, and sketch out its strategy. This gives you as the user a lot of power to make sure:
  1. Your business logic is understood
  2. The workflow will execute according to your specific needs
  3. There won’t be any data surprises.