🛒 DealSense AI

An intelligent agentic AI system that scans, evaluates, and discovers the best online deals in real-time. DealSense AI combines multiple AI models and specialized agents to predict product prices, identify significant discounts, and filter out irrelevant deals — ensuring users only see genuinely valuable offers.

✨ Core Features

• 🧠 Central Planning Agent: Orchestrates the complete workflow from deal scanning to price prediction and filtering
• 🔍 Real-time Deal Discovery: Fetches live deals from RSS feeds with automatic duplicate detection
• 💰 Advanced Price Prediction: Multi-model ensemble including OpenAI GPT, fine-tuned LLaMA, XGBoost, and RAG pipeline
• ⚡ Smart Filtering: Automated deal filtering based on configurable discount thresholds
• 📊 Transparent Operations: Complete logging of all agent actions and decision-making processes
• 🎯 Structured Output: Clean summaries with price estimates, discount percentages, and direct links
• 🧩 Memory System: Shared storage prevents processing duplicate deals
• ☁️ Scalable Execution: Compute-intensive models run remotely on Modal for optimal performance

🚀 Tech Stack

• Languages & Frameworks: Python, FastAPI, React
• AI & Machine Learning:
  • OpenAI GPT for deal selection and RAG-based price prediction
  • Fine-tuned LLaMA 3.1 8B (QLoRA) for specialized price estimation
  • XGBoost regression with E5 embeddings
  • RAG pipeline using ChromaDB vector database
  • Linear regression ensemble model

🏗️ Architecture

DealSense AI employs a multi-agent architecture with specialized components:

🤖 Specialized Agents

• Deal Scanner Agent: Fetches and filters deals from RSS feeds using OpenAI GPT
• Fine-Tuned LLM Agent: Price prediction using custom-trained LLaMA model on Modal
• RAG-Based Agent: Contextual price estimation with retrieval-augmented generation
• XGBoost Agent: Regression-based predictions using product embeddings
• Ensemble Agent: Combines all predictions for final price estimates

📊 Models & Data

• Dataset: 409K curated items from Amazon Reviews 2023 across 8 product categories
• Embeddings: E5-small-v2 model with ChromaDB for similarity search
• Training: Balanced sampling with price-based stratification
• Storage: Models and datasets hosted on Hugging Face Hub

🛠️ Setup

1. Environment Variables

   OPENAI_API_KEY=your_openai_key
   MODAL_TOKEN_ID=your_modal_token_id
   MODAL_TOKEN_SECRET=your_modal_token_secret

📈 How It Works

1. Scan: Deal Scanner Agent fetches real-time deals from RSS feeds
2. Predict: Multiple AI agents generate independent price predictions
3. Ensemble: Linear regression model combines predictions for final estimate
4. Filter: Deals below discount threshold are automatically filtered out
5. Present: Users receive structured summaries of valuable deals only

🎯 Use Cases

• Smart Shopping: Automatically discover genuinely discounted products
• Price Monitoring: Track market prices across multiple categories
• Deal Validation: Verify if advertised discounts represent real value
• Market Research: Analyze pricing patterns and trends

📖 Documentation

For detailed technical documentation, architecture deep-dives, and implementation guides, visit:

• Models & AI Components - Fine-tuned LLaMA, XGBoost, RAG pipeline, and ensemble details
• Features & Agents - Complete guide to all specialized agents and their capabilities
• Dataset & Curation - Data sourcing, filtering, sampling, and embedding processes
• API Integration - Internal API usage, authentication, and service integrations

---

DealSense AI leverages cutting-edge AI to transform how people discover and evaluate online deals, ensuring every recommendation represents genuine value.