https://github.com/xoxoharsh/hackdata-2025
https://github.com/xoxoharsh/hackdata-2025
Last synced: 3 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/xoxoharsh/hackdata-2025
- Owner: XoXoHarsh
- Created: 2025-03-08T08:09:23.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-03-08T09:52:45.000Z (3 months ago)
- Last Synced: 2025-03-08T10:23:37.458Z (3 months ago)
- Language: TypeScript
- Size: 257 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# LifeStyle – AI-Powered Scalable Health Assistant
*A lightweight, scalable AI chatbot for proactive health and lifestyle management.*
## Project Overview
LifeStyle is an AI-powered mobile application designed to seamlessly integrate healthcare and lifestyle management.
Instead of relying on large, resource-intensive LLMs, it uses multiple small, fine-tuned LLMs, each specialized in:
- Exercise & Fitness
- Mental Well-being
- Diet & Nutrition
- General Questions
The chatbot efficiently routes queries to the appropriate LLM using a classifier, making responses faster, accurate, and more affordable.
### Key Features
- **Lightweight and fast** – Uses small, task-specific LLMs instead of massive AI models.
- **Scalable microservices architecture** – Built with FastAPI and Docker.
- **Real time Alerts** - Real time alerts to get you to stop from over using unhealthy apps.
---
## Tech Stack
### Frontend (Mobile App) – React Native
- **React Native (TypeScript)** – Cross-platform mobile framework
- **Expo (TypeScript)** – Simplifies React Native development
- **StyleSheets** - To style the components.
### Backend – Python (FastAPI + Microservices)
- **FastAPI** – High-performance API framework
- **Docker** – To run Containerized model
- **Node** - Authentication and metric collection
- **FireBase FCM** - Real time notifications and alerts
### AI & Machine Learning
- **Ollama** – Running small LLMs locally
- **Fine-tuned llama3-8b** – Custom-trained on health topics
- **Unsloth** - For finetuning the llama3-8b and tinyllama model.
---
## Project Structure
```bash
Hackdata-2025
├── LifeStyle_App/ # React Native mobile app
│ ├── app/
│ ├── assets/
│ ├── utils/
│ ├── cache.ts
│ ├── package.json
│ ├── ...
├── Chat_Server/ # FastAPI Server
│ ├── Datasets/ # Synthectic Datasets used to fine-tune the models
│ ├── Training_Scripts/ # Scripts used to train models
│ ├── app.py # Main file to start the fastapi server
│ ├── requirements.txt
│ ├── ...
├── Mimic_Server/ # A server to handle stuff like a normal healthapp server
│ ├── src/
│ │ ├── controllers/
│ │ ├── db/
│ │ ├── middlewares/
│ │ ├── models/
│ │ ├── app.js
│ │ ├── index.js # Main file to start node server
│ │ ├── ...
│ ├── package.json
│ ├── ...
├── Model_Containers/
| ├── docker-compose.yml
| ├── Dockerfile # Scripts to run the models in containers
└── README.md
```
## Hugging face
``` sh
https://huggingface.co/XoXoHarsh
```
## How to run
``` sh
git clone https://github.com/XoXoHarsh/Hackdata-2025.git
```
## Chat_Server
```sh
pip install -r requirements.txt
pyhton app.py
```
### Traning Scripts
```sh
Datasets: The project includes four datasets—Classification_dataset.csv, diet_dataset.csv, excercise_dataset.csv, and stress_dataset.csv—used for training and classification.
Notebooks: Finetuning.ipynb consolidates multiple scripts for fine-tuning, and classification.ipynb handles text classification.
```
## LIfeStyle_App
``` sh
npm i
npx expo start
```
### env
```sh
EXPO_PUBLIC_CLERK_PUBLISHABLE_KEY=YOUR_CLERK_PUBLISHABLE_KEY
EXPO_PUBLIC_API_URL=YOUR_PUBLIC_API__URL
EXPO_PUBLIC_CHAT_SERVER=YOUR_PUBLIC_CHAT_SERVER_URL
```
## MIMIC SERVER
## Prerequisites
Make sure you have the following installed:
- [Node.js](https://nodejs.org/) (Latest LTS version recommended)
- [MongoDB](https://www.mongodb.com/) (Use MongoDB Atlas or local installation)
``` sh
cd Mimic_Server
cd src
nodemon index.js
````
## env
``` sh
PORT=8000
MONGODB_URI= YOUR_MONGODB_URL
ACCESS_TOKEN_SECRET=YOUR_ACCESS_TOKEN_SECRET
ACCESS_TOKEN_EXPIRY=2d
```
## App Screenshots
