AI Steam Game Price Forecasting

An advanced AI-powered system for analyzing and forecasting Steam game prices using Retrieval Augmented Generation (RAG) and Large Language Models.

🚀 System Overview

This project implements a sophisticated RAG-based system that combines real-time Steam data extraction with advanced language models to provide intelligent game price analysis and forecasting.

🏗️ Architecture

The system consists of four main components:

1. Data Extraction (data_extractor.py)

   • Selenium-based scraper for Steam's most played games
   • Real-time collection of game prices and player statistics
   • Automated CSV export with daily timestamps

2. Data Processing (data_loader.py)

   • Robust data preprocessing pipeline
   • Conversion to LangChain Document format
   • Type-safe implementation with error handling

3. LLM Integration (llm_client.py)

   • Custom Nvidia LLM implementation
   • LangChain-compatible interface
   • Configurable model parameters and streaming support

4. RAG Implementation (retrieval_chain.py)

   • FAISS vector store for efficient similarity search
   • HuggingFace MPNet embeddings
   • Advanced prompt engineering for accurate analysis
   • Retrieval-augmented generation chain

🛠️ Technical Stack

• Web Scraping: Selenium, Chrome WebDriver
• Data Processing: Pandas, NumPy
• Vector Store: FAISS
• Embeddings: HuggingFace sentence-transformers
• LLM Integration: Custom Nvidia API implementation
• RAG Framework: LangChain

🔧 Setup and Usage

1. Install dependencies:

pip install -r requirements.txt

2. Configure environment variables for API access

NVIDIA_API_KEY=your_api_key_here

3. Run the data collection:

python data_extractor.py

4. Remove NaN values:

python steam_game_price_analysis.py

5. Execute analysis queries through the RAG system:

python main.py

🎯 Key Features

• Real-time Steam data extraction
• Robust error handling and data validation
• Configurable RAG system with FAISS vector store
• Custom LLM integration with streaming support
• Type-safe implementation with comprehensive documentation

📊 Performance

• Efficient similarity search with FAISS
• Optimized embeddings using MPNet
• Streaming response generation
• Configurable retrieval parameters

📝 License

MIT License