ClickUp recently released an AI tool that allows users to have access to a chatbot, as well as other AI-powered features, that are based on the knowledge base that a specific project has on ClickUp. The user can ask the bot a question specific to their data, and it will search all the tasks, docs, and material that have been created inside of the user workspace. The bot can also provide suggestions and insights based on those data.
That left me wondering what the general process for achieving something like that with ChatGPT would be.
I’m not necessarily talking about the same type of functionality, but rather I’m referring to how GPT can help in achieving some type of AI chatbot which has a “common layer” (e.g. the ability to process and respond in natural language), and a second “layer” which can be customized according to a user-specific (or project-specific, or whatever) knowledge base.
Two scenarios I have hypothesized are:
- the model is fine-tuned for each tenant with the given knowledge base
- the model receives the whole knowledge base as context on each relevant prompt
Both of these seem unrealistic to me, but like I stated I don’t have a deep knowledge of this topic.
Can anybody enlighten me on this?