Introduction

Welcome to the documentation for the Yu-Gi-Oh! Ruling AI Chatbot project. This document provides an overview, setup instructions, and usage guides.

Key Features

• Accurate Ruling Responses: Provides precise answers for Yu-Gi-Oh! Goat Format rulings using a RAG approach.
• AI-Powered Chatbot: Combine AI reasoning and relevant context for comprehensive answers.
• Dockerized Environment: Ensures easy, consistent setup.
• Turborepo Monorepo: Efficiently manages project packages and applications.

Technologies Used

• Next.js: React framework for the chatbot.
• Vercel AI SDK: Integrates with AI models.
• OpenAI: Used for embeddings and AI models.
• Pinecone: Vector database for ruling embeddings.
• Tailwind CSS: Utility-first CSS.
• shadcn/ui: Reusable UI components.
• Turborepo: Monorepo management.
• ESLint & Prettier: Code linting and formatting.
• Cheerio: Web scraping library for server-side jQuery.
• Nextra: Documentation framework.
• Docker/docker-compose: Containerization.
• TypeScript: For type safety.

Retrieval-Augmented Generation (RAG)

This project uses RAG for accurate Yu-Gi-Oh! ruling answers. Instead of just using the AI’s general knowledge, RAG retrieves relevant ruling info from Pinecone’s vector database.

How it works:

1. Question Embedding: User question is converted to an embedding using OpenAI.
2. Retrieval: Embedding searches Pinecone for similar ruling embeddings.
3. Augmentation: Retrieved ruling text provides context to the AI model.
4. Generation: AI uses retrieved context and its reasoning to generate an accurate answer.

Here’s a fantastic blog  about RAG and why it’s used for more information.