Limiting the answers to specific questions

Hey,

We have an event platform which requires memberships.

We would like to implement an ai to our platform, so users can interact with it, ask questions or just talk…

The name should be “ask the universe”…

If the user asks about events / party’s (when is the next party, where is it etc) we would obviously want the api to only give answers regarding to our events…

And if they ask about the different memberships we want the ai to know the different pricings and advantages of the memberships…

Wich api would we need for that, and how would we prepare it so it knows the upcoming events and different memberships.

Thanks for your help.

@Anyframe - Welcome to the forums!

Personally, I would go with the Assistants API just because of the ease of implementation and control.

A good system prompt should work just fine. If you need help, you can ask in the Prompting category on the forums. You could also use function calling to get details from your API and send those outputs to the LLM for a response.

1 Like

I agree with Munna. You just need a system prompt like “You are a helpful assistant answering questions about events. Here is the relevant information for you to use…” then just keep explaining the relevant facts as sentences or bullet points, etc, that’s really all you need to do.

If there’s a vast amount of scheduled events or membership info (more than is appropriate for a System Prompt), you can either use RAG or Function Calling to make an agent be able to find the specific data for specific questions. But remember you can also replace the System Prompt with a new system prompt as you gather information during the conversation.

1 Like