MentalAI (work in progress)

MentalAI is a retrieval-augmented system for mental health support with fine-tuned suicide prediction. It leverages llama.cpp and LangChain to provide AI-assisted conversations based on retrieved contextual information.

Folder Structure

The repository is structured as follows:

mentalAI/
│── src/
│   ├── _0_data_preparation/       # Data processing pipeline
│   ├── _1_chroma_preparation/     # Setting up the vector store using ChromaDB
│   ├── _2_chat/                   # Chat system logic
│   ├── utils/                     # Utility functions
│   ├── __init__.py

Setup & Usage (More details follow soon)

Since the dataset for this project is private, users need to manually set up all required data resources. After preparing the data:

1. Set up the Chroma vector store following the scripts in src/_1_chroma_preparation/.
2. Run the chat system by executing:

   python src/_2_chat/chat.py

Technologies Used

• llama.cpp - Running LLaMA models efficiently on CPUs.
• LangChain - Managing AI-driven conversations and retrieval augmentation.
• ChromaDB - Vector store for efficient embedding retrieval.

Disclaimer

This project is intended for research and educational purposes. It is not a replacement for professional mental health services.