I’m looking for an AI to create a recommendation system based on a list of places that I will have on my Bubble application. It should be able to consider specific requests regarding the address, such as a neighborhood based on a text field filled in by the user. Do you have any suggestions for a service to use?
The client would write a brief in the form of text, and I would need to generate matches based on what they are requesting and the thousands of places that I have in the database.
What you’re looking to do can likely be accomplished with RAG if you already have a database of places. I might recommend using a service like carbon.ai which handles most of what you’re looking for if you’re not familiar with vector dbs. You just need to provide a query via an API call and it will return embeddings from the database you uploaded.
Your workflow would essentially be
Uploading your data to carbon via API
Construct a query and send it via API to carbon
Carbon matches your query with a list of embeddings which represent your data and returns that to you
Return that data back to the user as possible places
That would depend on the parameters which you specify. Most AI platforms which allow users to upload their own data are using some form of what I described. If you are looking for an alternative, you could try using SERP API which would give you as up-to-date results as it can be.
You can even go further and feed the results of the SERP into an OpenAI call with specifications for what the user is looking for.