Transcript
We can start by appending our description here. This becomes particularly interesting when we have a select field for different categories of vendors. We can set up an AI agent where we simply describe a supplier and then activate the agent to automatically fill the correct category value based on that vendor description.
To set this up, we add a prompt such as the vendor category for this and then tag the vendor description field. From there, we select a model like GPT-4o mini, since this is a relatively simple task. We can also enable web search if we need to look for external information about a specific company, though in this case, we are basing it solely on the provided description.
[.blog-callout]
While setting up these database agents manually is powerful, you can also use the AI co-builder to generate entire app structures and Softr Databases schemas that are already optimized for your specific data needs.
[.blog-callout]
There are settings to control when this AI agent runs. It can be triggered manually, whenever a new record is created, or when an existing record is updated. We can configure it so that whenever the vendor description field is updated, the agent re-triggers automatically.
Once we paste the description or update the field, the agent will run and go through the list of available values to handle the categorization. This is known as a database AI agent, and it gives significant superpowers to your database by allowing it to refine, clean, enrich, or reformat data ongoingly.
[.blog-callout]
Using these agents within Softr Databases is a great way to handle project management tasks like task triaging or status updates without needing external automation tools.
[.blog-callout]
If you have a field containing a PDF, the agent can also extract specific elements from that document to store as text. For example, it could extract an amount from an invoice or categorize the document type. Having AI living right inside your database significantly improves the overall efficiency of your app.
