Multi-Tenant LLM-powered FAQ Chatbot Platform

Project description

How can the architecture of a multitenant LLM-based Chatbot system, which will address Frequently Asked Questions (FAQs) using different and modifiable knowledge bases, be designed?

Context

CY2, an Oracle development partner, specialises in delivering software solutions to schools worldwide. They've previously used intent-based chatbots with Oracle's Digital Assistant frontend but found them slow to become operational. Now, they're considering using Large Language Models (LLMs) to answer FAQ questions.

The objective of the project is to provide CY2 with a software proof-of-concept (POC) multi-tenant software platform. With this, CY2 will have a flexible FAQ system driven by LLMs (Large Language Models) that allows their clients to create chatbots with a knowledge base based on their own documents. This can improve their present Oracle digital assistant chatbots by streamlining, eliminating manual training and requiring less input.

Results

We have developed a multi-tenant proof-of-concept solution that enables users to create their own chatbots. This solution leverages a user's unique knowledge base, which can be derived from PDF documents uploaded to data sources such as Microsoft SharePoint. Our platform utilizes advanced technologies like Langchain and Pinecone, and is built on a robust backend developed in NestJS. Additionally, we have designed a intuitive frontend to ensure an optimal user experience. This functional solution demonstrates the potential of generative AI in creating customized chatbot platforms.

About the project group

We are all Advanced Software semester students from all around Europe, all interested in using cutting edge technologies, hence our choice of group project.