I am working on a chatbot for my application where each user will log in and provide profile information (location, color preferences, etc.). This chatbot will be connected to the OpenAI API for handling responses to queries, and I plan to pre-train the AI with my own dataset for domain-specific knowledge.
How can I ensure that each user receives personalized responses based on their profile information and preferences, in addition to the domain specialization trained into the AI?
Furthermore, is it possible to link this setup to a payment processing system to facilitate transactions within the chat?
The simplest way is to implement your own
Custom Instruction function, specially the part pertaining to the user.
I’m not sure to understand what do you mean ?
So, for each request I will do to the API, I will add custom instruction in it according to the user connected ?
I made some role-playing chatbot before. In the system prompt, there are two parts: one instruction about the AI and another instruction for who the user is. This is basically, how
Custom Instruction works.
For example, your app is a role-playing app for LOTR: the first part of the system prompt, you tell the AI to act as Gandalf, then the second part, you tell the AI that you are Frodo, if the user selected to be Frodo.