Righthere - AI-Powered Diary Analysis App

Righthere is a .NET MAUI application that leverages artificial intelligence to analyze diary entries and provide mental health advice. It combines machine learning and AI technologies for emotion detection and therapeutic feedback.

🤖 AI Features

1. Emotion Detection

• Model: Custom fine-tuned model on Hugging Face Hub
• Model Repository: Gun555/Righthere
• Base Model: Pre-trained transformer for emotion classification
• Capabilities: Detects 6 emotion types
  • 😢 Sadness
  • 😊 Joy
  • ❤️ Love
  • 😠 Anger
  • 😨 Fear
  • 😲 Surprise

2. AI Therapist

• AI Engine: Google Gemini 2.0 Flash
• Role: AI Therapist for emotional analysis and advice
• Response Structure:
  • Suggestion: Positive and supportive advice
  • Emotional Reflection: Summarizes and analyzes user emotions
  • Mood: Assigns a primary emotion in one word
  • Keywords: Extracts key topics from the diary entry

🚀 Setup Instructions

Prerequisites

• .NET 9.0 SDK
• Python 3.8+ (for API backend)
• Google AI Studio API Key
• Hugging Face Token (for model access)
• Internet connection (for downloading model from Hugging Face Hub)

1. AI Model Setup

The model will be automatically downloaded from Hugging Face Hub on first run:

• Model Repository: https://huggingface.co/Gun555/Righthere
• Auto-download: Model will be cached locally after the first download
• No manual download required

2. Environment Configuration

Create a .env file in the api/ folder with the following variables:

# Google Gemini API Key (get from Google AI Studio)
apikey=your_google_ai_studio_api_key_here

# Model path (if using a local model)
MODEL_PATH=path/to/your/model/folder

# Hugging Face Token (for accessing Gun555/Righthere model)
HuggingFaceToken=your_huggingface_token_here

# Server Port (optional)
PORT=8000

3. Technical Architecture

User Input (Diary) → Hugging Face Model Hub → Emotion Detection → AI Analysis → Therapeutic Response
                             ↓                       ↓                ↓
                    Gun555/Righthere      →  Transformers Pipeline  →  Google Gemini API  →  Structured Advice

4. Model Details

• Model Type: Sequence Classification (Text → Emotion Labels)
• Framework: PyTorch + Transformers
• Input: Thai/English text
• Output: Probability distribution over 6 emotion classes
• Performance: Optimized for diary/personal text analysis

5. Install Python Dependencies

cd api
pip install -r requirements.txt

6. Run the Application

Start AI API Backend:

cd api
python app.py

🛠️ API Endpoints

/getadvice (POST)

{
  "text": "Your diary entry text"
}

Response:

{
  "emotion": "joy",
  "advice": "AI-generated advice and emotional analysis"
}

/ (GET)

Health check endpoint

🎯 Model Information

Hugging Face Model: Gun555/Righthere

• Model Card: https://huggingface.co/Gun555/Righthere
• Model Type: AutoModelForSequenceClassification
• Tokenizer: Compatible with Transformers AutoTokenizer
• Input Format: Text sequences (Thai/English)
• Output: 6-class emotion classification

Model Performance

• Classes: 6 emotion categories (sadness, joy, love, anger, fear, surprise)
• Optimized for: Personal diary and emotional text analysis

Using the Model Directly

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load model and tokenizer
model_name = "Gun555/Righthere"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Predict emotion
text = "I feel great today!"
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=128)
outputs = model(**inputs)
predicted_class = torch.argmax(outputs.logits, dim=-1)

🔧 Development

Technology Stack

• Frontend: .NET MAUI (C#)
• Backend: FastAPI (Python)
• ML Framework: Transformers (Hugging Face)
• AI Model: Custom model hosted on Hugging Face Hub (Gun555/Righthere)
• AI Service: Google Gemini API
• Model Hosting: Hugging Face Model Hub
• Database: Local storage with Models/

Project Structure

Righthere/
├── api/                 # AI Backend API
├── Models/             # Data models (C#)
├── Services/           # API services (C#)
├── Views/              # MAUI pages
├── Platforms/          # Platform-specific code
└── Resources/          # App resources