RAG ChatBot with Chat History

A conversational RAG (Retrieval Augmented Generation) system that remembers your conversation history and provides contextual answers from documents.

🎯 What It Does

Transform any document into an intelligent chatbot that:

• Remembers conversations: Maintains context across multiple questions
• Answers from documents: Retrieves relevant information to answer your questions
• Interactive chat: Real-time conversation with command support
• Persistent storage: Saves chat history using Chroma vector database

🚀 Quick Start

1. Install dependencies:

pnpm install

2. Set up environment:

cp env.example .env

3. Add your API keys to .env:

OPENROUTER_API_KEY=your_openrouter_key_here
OPENAI_API_KEY=your_openai_key_here

4. Start chatting:

# Interactive chat mode
pnpm start --interactive

# Sample questions mode
pnpm start

# Streaming answers
pnpm start --streaming

💬 Usage Examples

Interactive Chat

$ pnpm start --interactive

💬 You: What is task decomposition?
🤖 Assistant: Task decomposition is the process of breaking down complex tasks into smaller, manageable steps...

💬 You: Can you give me an example?
🤖 Assistant: Based on our previous discussion about task decomposition, here's an example...

Available Commands

• /help - Show available commands
• /reset - Start a new conversation
• /history - View conversation history
• /threads - Show all conversation threads
• /switch <thread_id> - Switch to different conversation
• /summary - Generate conversation summary
• /exit - Exit the chat

🏗️ Architecture

The system uses a conversational StateGraph approach:

User Input → Query Reformulation → Document Retrieval → 
Contextual Response → Chat History Storage

Key Components:

• Chroma Vector Database: Stores documents and embeddings
• LangGraph StateGraph: Manages conversation flow
• Chat History Manager: Handles conversation persistence
• OpenRouter Integration: LLM and embedding models

📁 Project Structure

rag-langchain/
├── src/
│   ├── wrappers/
│   │   ├── chroma-wrapper.js       # Chroma database wrapper
│   │   └── embeddings-openai.js    # OpenAI embeddings wrapper
│   ├── chat-history.js             # Conversation management
│   ├── interactive-chat.js         # CLI chat interface
│   ├── rag.js                      # Main RAG system
│   └── config.js                   # Configuration
├── index.js                        # Entry point
└── README.md

⚙️ Configuration

Edit src/config.js to customize:

• Models: Change LLM and embedding models
• Chroma Settings: Database configuration
• Chat History: Conversation persistence options
• Prompts: System prompts in Korean/English

🔧 Use Cases

1. Document Q&A Chatbot

• Load company documents, manuals, or knowledge bases
• Create an interactive assistant that answers questions
• Maintain conversation context for follow-up questions

2. Customer Support Assistant

• Upload FAQ documents and product manuals
• Provide contextual customer support
• Remember previous conversation for better assistance

3. Research Assistant

• Load research papers or articles
• Ask complex questions with follow-ups
• Generate summaries of long conversations

4. Personal Knowledge Base

• Store personal documents and notes
• Create a conversational interface to your knowledge
• Search and retrieve information naturally

🎨 Customization Examples

Add New Document Loaders

// Add PDF loader
import { PDFLoader } from "@langchain/community/document_loaders/fs/pdf";

// Add to RAGSystem class
async loadPDF(filePath) {
  const loader = new PDFLoader(filePath);
  return await loader.load();
}

Custom Chat Commands

// Add to interactive-chat.js
case '/export':
  await this.exportConversation();
  break;

Connect to External Chroma Server

CHROMA_USE_LOCAL_DB=false
CHROMA_HOST=your-chroma-server.com
CHROMA_PORT=8000

🛠️ Models

Default Models:

• LLM: moonshotai/kimi-k2:free (via OpenRouter)
• Embeddings: text-embedding-3-small (via OpenAI)

Change Models:

LLM_MODEL=anthropic/claude-3-haiku
EMBEDDING_MODEL=text-embedding-3-large

🐛 Common Issues

"Better SQLite3 bindings not found"

• Conversation persistence will fallback to memory storage
• Install build tools: pnpm add --dev node-gyp

"Chroma client initialization failed"

• Using memory-based vector store as fallback
• Works normally with in-memory storage

"API key issues"

• Ensure both OpenRouter and OpenAI keys are set
• Check API key validity and credits

📚 Advanced Features

Conversation Management

• Thread Support: Multiple conversation threads
• Persistence: SQLite-based chat history storage
• Context Awareness: Query reformulation based on chat history
• Summarization: Auto-generate conversation summaries

Vector Database

• Chroma Integration: Production-ready vector storage
• Similarity Search: Efficient document retrieval
• Embeddings: High-quality text embeddings
• Fallback: Memory storage for development

🎉 Getting Started Tips

1. Start with interactive mode to experience the conversational flow
2. Try follow-up questions to see context awareness in action
3. Use /help to explore available commands
4. Experiment with different document URLs in the config
5. Check conversation history with /history command

---

Ready to chat with your documents? 🚀

pnpm start --interactive