I have a knowledge base (description of users and companies) stored in a chromadb vector database. Then I have a relational database where I store user interaction in my app (which and how many times profiles he visited, texts he read, videos he played, profiles he commented). I am planning to use OpenAI to make recommendations to a user based on both stored knowledge and his own previous interaction.
Why I want to do that? I want suggest company profiles to user, he may like.
What I do know how to do? Knowledge base works perfectly. Texts about user are matched against company descriptions and the most suitable companies are found for user.
What I don't know how to do? The question is how to provide that user interaction data to OpenAI in order to OpenAI will understand it, since the data are just relations and numbers. I was thinking about transforming the data into following sentences...
- "I am interested into 'name of article' and visited it 'count' times"
- "I visited profile of company 'name of company' 'count' times and interacted with contact form"
- "I visited profile of company 'other company name' once and clicked that i don't want to receive notifications from this company anymore."
And pack these sentences as an additive context for OpenAI query. But is this correct approach? Can you suggest me a right / optimal solution for this?