Hello, I am quite new to using OpenAI’s API.
I am interested in developing a chatbot service that utilizes my private dataset for responses. My initial plan was to use OpenAI’s API for facilitating conversations between the chatbot and users.
However, I understand that the Assistant AI is primarily designed for incorporating private datasets, mainly by retaining context. My concern is that this focus on context retention might not align perfectly with my requirements, as my application does not need to maintain context.
Is there an alternative method within OpenAI’s framework that allows my chatbot to use my private dataset?
The assistants agent framework may make it easier, but it is nothing that didn’t come before, implemented by independent API developers to perform similar tasks. In fact, “assistants” being a general-purpose kind-of-do-what-you-need idea, can be surpassed in function and cost in almost any situation by development that uses direct access to the AI models.
The primary method one would employ to make an AI smarter about your own knowledge is what has become known as retrieval-augmented generation (RAG). The user inputs a query, and an AI-powered search is done against a knowledge database that automatically puts some top results into the AI’s context memory so it can answer better.
The string of keywords for building this: embeddings-based semantic search vector knowledge chunked database.
I’ve already shared this a couple times, so I’ll share it again in case you need to come up to speed on RAG: I want to create a continuously improving legal AI - #5 by SomebodySysop
Is it OK if a moderator closes this topic?
Yes, it’s ok to close. Thanks!