IntraMind

IntraMind AI Chatbot Integrated Platform

Inspiration

We're inspired by real demand in real industries.

One day, my fitness coach (founder and CEO of his fitness center), knowing that I study computer science, discussed his next business plans with me. As the macro-level economic environment stripped down aggregate consumption, many cheaper gyms shut down. However, since there're always the richer guys in any societal environment who would be able to maintain the way they live no matter what, and since my coach's gym mainly targets the group of people with higher consumption abilities, he now planned to broaden his business and try to take in the demands that other fitness centers have lost. He has much better knowledge and experience than most of the fitness coaches out there, so he plans to put his knowledge into AI knowledge base, which would allow him to train the newbie coaches to someone like him with a shorter period because those coaches would just need to enter the data reports of their clients and they'll get a step-by-step guide of doing exactly what my coach would've done.

I know it's hard nowadays to venture into something unknown, so I was really touched when I heard the plan. From then on, I know I will build this for him.

What it does

IntraMind is a Retrieval-Augmented Generation (RAG) AI Chatbot driven by Streamlit that allows users to login to and manage their own knowledge base by uploading files and chat with the bot to get more accurate, reliable, and secure answers about any private fields that basic LLMs cannot answer.

How we built it

We have two engineers in total, both of them freshmen studying in either CS or ECE. We split our responsibilities into frontend and backend. The frontend would explore Streamlit, UI design, OAuth, chat and UI workflows, etc., and the backend would implement database and table structures, APIs that satisfy frontend needs, inner logics and algorithms of embeddings, chat functionalities, and management of session histories. We have fixed weekly meetings and on-call temporary syncs just like what engineers in real industries would do.

Challenges we ran into

Although we're both strong in logical reasoning and have sufficient programming experience, we're both new to many of the more detailed usage of implementing a RAG AI Chatbot that has to be adjusted to satisfy various personalized functionalities. We sometimes ran into version discrepancies or environment issues that we haven't expected, but we debugged them one by one and we learned a lot from that experience as well.

What's next for IntraMind

We built IntraMind within a very limited amount of time, and we're both doing heavy school work along with other projects, so the current status of IntraMind is still not something we would ultimately stop at. In the future, the following can be promised:

We'll make the frontend appearance to be more appealing.
We'll allow user interaction. Each user can manage many different knowledge bases, and users can visit other people's knowledge bases via link/key shared.
We'll put it on a website so that every user can access it more conveniently.
We'll improve the processing/rendering speed.

Hope you'll enjoy it. Stay tuned~