I’ve been playing with interfacing Knack with OpenAI and have to say I’m very impressed. My use case is risk management - I’ve been feeding OpenAi the risk, causes, consequences and controls for a risk, and asking it to provide any additional potential causes, consequences and controls. Very useful responses.
BUT.
My challenge is that risks, causes, consequences and controls are all in related tables, and other than creating a report I have no way to capture all of these into a “payload” to send to OpenAI.
Right now I’m cheating and producing a report, which lists the risk, causes and controls, consequences and controls, and then grabbing that text and adding it manually to a field that I then use as the payload for the API call to OpenAI.
But of course that doesn’t fulfil the need.
Does anyone have any thoughts on how to automatically create a field that is the concatenation of a bunch of different fields from across a set of related tables?
You could write a small amount of JS for each record update event that happens from these related records and consistently write those updates to fields in a table that will serve as your “flat” structure — then use a text formula to concatenate those fields in the main table where you’re updating these fields.
You might be limited depending on how much data you’re storing in each field, but the principle of storing data per record update where you want it can help you consolidate the data you need for the prompt with OpenAI.
Could you create a Questions table that uses text equations to concatenate the existing tables? You could even have a field in the table to receive the answer from your AI Model.
Or maybe I misunderstand the situation.
Use text function to create one field with all the information you need from one table or related one. also you can use html tags to style it if you would like to show it in details view.