System Block Diagram

Description

The flow of this project starts with the user interacting with a frontend. Then, through that frontend, the user is able to interact with the Keystone bot. This bot interacts heavily with an LLM (Large Language Model) in order to supply information that the user is interested in. The LLM is used to help parse the data that the user may be interested in and narrow down the useful documentation from the superfluous. The logic connects it to a database specific to the organization that is using the frontend. This database is where the organization specific data and files are located - files such as notes on common tools or completed projects that the team has worked with. It also has a subsection that holds information that the Keystone project knows about its users: information about seniority and a suspected knowledge base (the bot learns from previous interactions and asks the user to rate their comfortability with the topics) that the bot can then use to help build bridges between coworkers and reference them to each other. Then, the LLM is fed the important documents in order to provide the user with the most apt information possible. If the information in the database is insufficient, it may look to the internet in order to find papers and documentation that might help the user instead. Then, the information collected by the Keystone bot is sent through the frontend to the user.