EmoSApp: Offline Conversational Agent for Mental Health Support on Smartphone

Overview

Welcome to EmoSApp, an innovative, entirely offline conversational application designed to provide emotional and mental health support directly on your smartphone. This app leverages the power of Large Language Models (LLMs), specifically a fine-tuned and quantized version of LLaMA-3.2-1B-Instruct. EmoSApp ensures all AI inference happens locally on your device, prioritizing privacy and accessibility.

Key Features

100% Offline: All inference happens on-device, with no internet access required, making support available anytime, anywhere.
Enhanced Privacy: All conversation logs are stored locally on the user's device, safeguarding sensitive mental health data.
Resource-Efficient: Employs quantized models for responsive performance on modern smartphones, even for those with as low as 6GB of system RAM.
Empathetic & Coherent: Delivers understanding and contextually relevant responses, maintains interactive dialogues, and offers helpful suggestions for mental well-being.

Repository Contents

EmoSApp.apk: Android application package for EmoSApp.
Full_fine_tuned_model.pte: Contains the Full fine-tuned LLaMA-3.2-1B-Instruct model.
QAT-LoRA_fine-tuned.pte: Contains the QAT-LoRA fine-tuned LLaMA-3.2-1B-Instruct model.
tokenizer.bin: The tokenizer used by both model variants.
Knowleadge_dataset.json: A curated dataset containing 14,582 question–answer pairs grouped into 1,459 conversational QA dialogues.

Setup and Installation for the EmoSApp

Install the APK:
- Download EmoSApp.apk from the directory.
- Tap the APK file to install. Install it on your Android smartphone.
Load Model & Tokenizer:
- Open EmoSApp.
- Access the settings menu (From the top bar).
- Tap "LOAD MODEL". This will load the model and tokenizer into your device's memory.
- Wait for “Successfully loaded model” confirmation.
Begin Conversation:
- Once loaded successfully, you can start the conversation with our EmoSApp.