!!! Note: This project is currently in active development. Features and documentation may change as the project evolves.

BookTalk RAG Chat

A smart documentation assistant that transforms any GitBook documentation into an interactive AI-powered chat interface. Built with low-code tool Retool.

🚀 Features (WIP...)

• Multi-Source Documentation: Ingest GitBook documentation from any public URL.
• Intelligent Chat Interface: Ask questions in natural language and get contextual answers
• Multi-LLM Support: Choose between chat API
• Vector Search: Semantic similarity search for more accurate information retrieval

🛠️ Tech Stack

• Frontend: Retool (Self-hosted)
• Database: PostgreSQL with pgvector extension
• AI/ML: Bert Embeddings, Multiple LLM APIs
• Infrastructure: Docker Compose
• Vector Storage: pgvector for semantic search

🚀 Quick Start

Prerequisites

• Docker and Docker Compose installed

1. Clone and launch:

git clone git@github.com:nickgore/BookTalk.git
cd BookTalk
cp .env.example .env
# !!! Edit .env with your valid retool license key

# Launch all services
docker-compose up -d

Note: Get a free Retool license at retool.com for self-hosted deployment

2. Initial Retool Setup

1. Open Retool in your browser:

   http://localhost:3000/auth/signup
   

2. Create admin account:

   • Fill in your email address
   • Create a secure password
   • Complete the setup wizard

3. You'll be redirected to the Retool dashboard

3. Import the BookTalk App

• Click "Create" → "From JSON/ZIP"
• Upload the file: ./retool/apps/BookTalk.json

🏗️ Workflow

GitBook URL → Content Scraper → Text Chunker → Embedding Generator → Vector Database
                                                                             ↓
User Question → Vector Search → Context Retrieval → LLM API → Response with Citations

🔮 Roadmap

• Phase 1: Basic URL processing and chat interface
• Phase 2: OpenAI API integration
• Phase 3: GitBook parser with API interface (python, langchain)
• Phase 4: Vector search
• Phase 5: Multi-LLM support and advanced retrieval

Monitoring with Portainer

Access Portainer at http://localhost:9000 to monitor your containers in real-time. You can view resource usage, logs, and restart services if needed.