I would like to know if it is possible to help me because I want to create a small application to add to the portfolio using Openai applications. The idea is to make a chatbot assistant or customer service for the restaurant that my mother manages, in this way I would be applying it to a real scenario and it would serve me to acquire knowledge.
The idea is that the chatbot can answer questions and give information about the restaurant, its history, special menus and so on. Additionally, it should be able to make a request to an api using the Openai functions of the gpt4 and gpt3.5 models. In this case the information must be the necessary to create a ticket in the api with the customer and his order.
I don’t know if I made myself understood, I remain attentive anyway and thank you very much.
This is certainly possible, and is one of the most common ways to use language model AI.
You can give specific instructions to an OpenAI engine that it will only answer questions from a specific domain, such as being the representative of your restaurant.
You can provide it extended information that allows it to answer a question from a user about your restaurant by setting up a database retrieval system that gives the AI the information it needs before it provides an answer. This type of system is called an embeddings vector database. You can look at “openai cookbook vector databases” as a starting point to some of the systems that can be loaded with a custom knowledge base and retrieve information pertinent to the most recent line of questioning from the client.
The operation of the database can be by function calling features added to the lastest versions of chat models. By simply specifying a function such as “restaurant menu search”, you can write an external search that could provide further information on demand when the AI wants it.
I wish you success in your exploration of this new technology!