Help Me!

Basic questions that can be asked
More questions that can be asked to the chatbot
Simple log in window

Inspiration

We wanted to create a product that would help people, especially with those who have trouble finding some. This project makes it easier for those who need help, find some.

What it does

It has two core functionalities:

Have special chatbot who users can chat with. Users would write down how they are feeling, what those feelings mean for them, and empathize with the bot. The bot would just “listen” to the user. That way, the user feels like they are understood, and receive some empathy. Users would make these journal entries and chat with the bot. The bot only asks questions to get the user to talk more about how they are feeling and being affected
To connect users with mental health issues with therapists. Therapists will only have access to user chatbot history when the user selects the therapist they want. This speeds up the time for the therapist to understand the user and their problems, and how they can be helped. Also, if multiple users have similar issues, the therapist can create a “emotional circle” for all the users, where they once again can empathize together, and have that feeling where someone understands them.

How we built it

Our project is composed of a few frameworks and libraries. We made use of Django (hence, it is all written in Python), Firebase (which gives us a NoSQL database), and for the chatbot, we made use of vector databases and open source LLM models.

Challenges we ran into

Finding datasets for our chatbot. Running LLMs on hardware and resources. Sometimes, the bot could run on specific platforms only.

Accomplishments that we're proud of

We are most proud of creating the chatbot, and how it can ask open ended questions and empathize with the user. We are also proud of how much we learned during development.

What we learned

We learned about the strengths and weaknesses of multiple LLMs and just generally, Django. We also learned different vector database implementations.

What's next for Help Me!

Creation of hyper tuned LLMs that disorders and symptoms. We also want to make a special "immediate help" feature for those who are feeling very down and need someone to talk to. Also improve the dataset for the LLM by adding more journals, and other data to learn from.

Built With

Submitted to

UWB Hacks: Save The World!

Created by

I integrated the Django framework. This includes writing up the code for the dashboard, and processing requests (for instance, getting data from the chatbot) on the backend

Michael Shapiro
I developed the Retrieval-Augmented Generation (RAG) system by integrating a vector database with an open-source LLM. I learned a lot about the scaling considerations for large language models and techniques for augmenting model responses with relevant and accurate information.

Private user