Bootcamp/Euclid (Chijioke Nna): Week 5 - Video Game Trivia RAG System

[cjayprime]

PR: Add Video Game Trivia RAG Application (Game Master)

Overview

This PR introduces the Game Master, an interactive Retrieval-Augmented Generation (RAG) application designed to answer questions about video games using a curated knowledge base. The tool demonstrates core RAG concepts through an engaging gaming theme, utilizing vector embeddings and semantic search to retrieve and present relevant game information through a sleek Gradio interface.

Key Features

• Vector Database Integration: Implements ChromaDB for efficient similarity search, converting game information into searchable embeddings using SentenceTransformers (all-MiniLM-L6-v2 model).
• Semantic Search Engine: Transforms user queries into vector embeddings and retrieves the most relevant game documents based on semantic similarity rather than just keyword matching.
• Multi-Tab Gradio Dashboard: A feature-rich interface with four specialized tabs:
  • Ask Questions: Main Q&A interface with example prompts
  • Compare Games: Side-by-side game comparison functionality
  • Random Facts: One-click random interesting fact generator
  • Knowledge Base: Browse all games in the database
• Context-Aware Response Generation: Dynamically formats retrieved information based on query intent (e.g., highlighting release dates for "when" questions, developer info for "who" questions).
• Interactive Learning Demo: Perfect for understanding RAG workflows with immediate visual feedback on retrieval quality.

What’s Inside

• Backend: Vector store using chromadb with document chunking and metadata tagging for precise retrieval.
• Logic: Custom retrieval function with relevance scoring, query intent detection, and dynamic response formatting without requiring an LLM API.
• Data Layer: Curated JSON knowledge base of 5 popular games with rich metadata including release years, developers, genres, descriptions, and interesting facts.
• Visuals: Clean Gradio interface with responsive design, example galleries, and formatted Markdown outputs for readable responses.
• Workflow: Complete Jupyter notebook execution flow from data ingestion → vector embedding → Gradio deployment, all runnable locally.

---

Screens:

CC: @ranskills