rizzGPT

submissions/README.md

@@ -0,0 +1,173 @@
+# RAG-Enhanced Real-Time Dating Coach
+
+**AI@GT Applied Research Team Submission by John Doe**
+
+## Project Title
+RAG-Enhanced Real-Time Dating Coach - Generative Conversation Assistant
+
+## What It Does
+
+This application is an advanced real-time conversation coach that uses **Retrieval-Augmented Generation (RAG)** with embeddings to provide dynamic, contextual conversation suggestions. Instead of pre-coded templates, it generates authentic responses by retrieving and adapting examples from a large corpus of real conversations.
+
+### 🚀 **Key RAG Features:**
+- **Semantic Search**: Uses sentence transformers to find contextually relevant conversation examples
+- **Vector Database**: FAISS index with 5,000+ processed conversation examples for fast retrieval  
+- **Dynamic Generation**: Creates unique suggestions adapted to current conversation context
+- **Context-Aware Retrieval**: Considers sentiment, emotion, engagement, and topics for relevant examples
+- **Structured Output**: Generates coherent, natural suggestions based on retrieved conversations
+
+### 🔥 **Advanced AI Features:**
+- **Real-time sentiment analysis** using `cardiffnlp/twitter-roberta-base-sentiment-latest`
+- **Emotion detection** using `j-hartmann/emotion-english-distilroberta-base`
+- **Sentence embeddings** using `all-MiniLM-L6-v2` for semantic similarity
+- **RAG pipeline** with `the-rizz/the-rizz-corpus` dataset (55,000+ examples)
+- **Conversational lowercase style** for natural, casual suggestions
+- **Multi-layered context analysis** (sentiment + emotion + engagement + topics)
+- **Boundary detection** with respectful communication prioritization
+
+### What Makes It Special:
+- **🧠 RAG-Powered Intelligence**: Retrieves contextually relevant examples from thousands of conversations
+- **🎯 Dynamic Generation**: No fixed templates - generates unique suggestions every time
+- **📊 Semantic Understanding**: Uses vector embeddings to understand conversation meaning and context
+- **🔍 Context-Aware Retrieval**: Considers sentiment, emotion, engagement level, and conversation topics
+- **💬 Natural conversational tone** - suggestions feel authentic and human-like
+- **🚫 Never auto-sends** - all suggestions are for human consideration only
+- **🛡️ Boundary-aware** - includes advanced boundary detection and respectful communication
+- **🔒 Privacy-focused** - runs locally with FAISS vector database, no external API calls
+- **📚 Continuously learning** - adapts suggestions based on conversation corpus analysis
+
+## How to Run It
+
+### Prerequisites
+- Python 3.7 or higher
+- Internet connection (for initial model download)
+
+### Installation Steps
+
+1. **Clone or download the project:**
+   ```bash
+   cd submissions/John_Doe/
+   ```
+
+2. **Create a virtual environment (recommended):**
+   ```bash
+   python -m venv venv
+   # On Windows:
+   venv\Scripts\activate
+   # On Mac/Linux:
+   source venv/bin/activate
+   ```
+
+3. **Install required dependencies:**
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+4. **Run the application (recommended - RAG-enhanced):**
+   ```bash
+   python run_coach.py
+   ```
+
+5. **Or run directly:**
+   ```bash
+   python rag_dating_coach.py    # RAG-enhanced version (recommended)
+   python dating_coach.py        # Original version
+   ```
+
+### First Run Notes
+- The first run will download AI models (~500MB total) - this may take a few minutes
+- Models are cached locally for faster subsequent launches
+- The app works entirely on CPU - no GPU required!
+
+## How to Use
+
+1. **Start the application** - A GUI window will open
+2. **Paste conversation text** in the left panel (copy from your messaging app)
+3. **Click "Analyze Conversation"** or enable auto-analysis
+4. **Review AI suggestions** in the right panel
+5. **Use suggestions as inspiration** - adapt them to your authentic voice
+6. **Save analysis** for later review if desired
+
+### Example Usage (RAG-Enhanced):
+```
+Conversation Input:
+alex: hey! how was your weekend?
+sam: it was incredible! went rock climbing for the first time
+alex: no way! that sounds terrifying but amazing
+sam: right?? i was so scared at first but the adrenaline rush was unreal
+alex: i've always wanted to try that! where did you go?
+sam: there's this awesome indoor gym downtown, perfect for beginners
+alex: that's so cool! maybe i should check it out sometime
+sam: you totally should! i'd love to go again, maybe we could go together?
+
+RAG-Enhanced Analysis Output:
+sentiment: Positive (0.99)
+emotion: Joy (0.76)
+engagement: High
+rag system: ✅ Active
+
+🤖 rag-generated suggestions based on similar conversations:
+💬 that's so exciting! rock climbing sounds like such an adventure
+💬 i love how you stepped out of your comfort zone! that's really inspiring  
+💬 you're making me want to try it too! i'd definitely be up for going together
+💬 the fact that you faced your fear and loved it shows what an amazing person you are
+🔥 high engagement + shared activity interest: perfect opportunity for connection!
+✨ they're clearly excited and open to spending time together
+🎯 next move: express genuine interest and suggest a specific plan
+```
+
+## Technical Details
+
+### AI Models & RAG Architecture:
+- **Sentiment Analysis**: `cardiffnlp/twitter-roberta-base-sentiment-latest` - Robust Twitter-trained sentiment classifier
+- **Emotion Recognition**: `j-hartmann/emotion-english-distilroberta-base` - Multi-emotion classification
+- **Embeddings**: `all-MiniLM-L6-v2` - Sentence transformer for semantic similarity
+- **Vector Database**: FAISS with cosine similarity for fast retrieval
+- **Conversation Corpus**: `the-rizz/the-rizz-corpus` - 55,000+ conversation examples, filtered to 5,000 high-quality examples
+
+### RAG Pipeline Architecture:
+1. **Corpus Processing**: Clean and filter conversation examples for quality
+2. **Embedding Generation**: Create vector representations using sentence transformers  
+3. **Index Creation**: Build FAISS vector database for fast similarity search
+4. **Context Analysis**: Analyze current conversation for sentiment, emotion, topics
+5. **Retrieval**: Find most relevant conversation examples using semantic search
+6. **Generation**: Adapt retrieved examples to current context and generate suggestions
+7. **Filtering**: Apply safety filters and context relevance scoring
+
+### Safety Features:
+- **Boundary Detection**: Scans for keywords indicating discomfort or disinterest
+- **Respectful Templates**: Pre-vetted response suggestions that avoid toxic patterns
+- **Context Awareness**: Analyzes recent message history rather than isolated responses
+- **No Auto-sending**: Human always remains in control of actual communication
+
+### Privacy & Ethics:
+- **Local Processing**: All analysis happens on your device
+- **No Data Collection**: Conversations are not stored or transmitted
+- **Consent-Focused**: Emphasizes respectful communication and boundary recognition
+- **Educational Purpose**: Includes tips for healthy relationship communication
+
+## File Structure
+```
+submissions/John_Doe/
+├── rag_dating_coach.py      # RAG-Enhanced main application (RECOMMENDED)
+├── dating_coach.py          # Original implementation for comparison
+├── run_coach.py             # Simple launcher script
+├── config.json             # Configuration settings
+├── requirements.txt        # Python dependencies
+└── README.md               # This documentation
+```
+
+## Future Enhancements
+- Voice analysis integration
+- Multi-language support
+- Advanced personality matching
+- Integration with popular messaging platforms
+- Mobile app version
+
+## Disclaimer
+This tool is designed to promote respectful, authentic communication. It should be used as a learning aid, not as a replacement for genuine human connection and emotional intelligence. Always prioritize consent, boundaries, and authentic self-expression in your relationships.
+
+---
+
+**Developed for AI@GT Applied Research Team Assessment**  
+*Focus: Hugging Face transformers, ethical AI, user interface design*

submissions/config.json

@@ -0,0 +1,31 @@
+{
+    "models": {
+        "sentiment_model": "cardiffnlp/twitter-roberta-base-sentiment-latest",
+        "emotion_model": "j-hartmann/emotion-english-distilroberta-base"
+    },
+    "analysis": {
+        "max_messages_to_analyze": 10,
+        "auto_analysis_delay": 2.0,
+        "min_message_length": 2
+    },
+    "safety": {
+        "boundary_keywords": [
+            "stop", "uncomfortable", "not interested", "leave me alone",
+            "busy", "can't talk", "not now", "please stop", "don't want to",
+            "feeling pressured", "need space", "not in the mood"
+        ],
+        "respect_filters": true,
+        "enable_boundary_detection": true
+    },
+    "ui": {
+        "window_width": 1000,
+        "window_height": 700,
+        "auto_analyze": false,
+        "save_history": true
+    },
+    "suggestions": {
+        "max_suggestions": 4,
+        "include_next_steps": true,
+        "personalization_level": "medium"
+    }
+}