🧠 ChatBotForge – Custom AI Chatbots for Businesses

ChatBotForge is a platform that allows businesses to create AI-powered chatbots trained on their own data — including documents, websites, and internal knowledge. Each bot is isolated, customizable, and easily embeddable on any site.

---

🚀 Features

For Businesses

• Create and manage multiple bots
• Upload files (PDF, DOCX, TXT) or crawl web pages
• Fine-tune bot tone, greeting, and settings
• Test bots instantly in a chat playground
• View chat history and analytics (optional)

For End Users (Your Customers)

• Real-time AI chat interface
• Context-aware responses based on business content
• Multi-language support
• Mobile-friendly UI
• Easy embed via iframe, widget, or SDK

---

🧱 System Architecture Overview

Components

• Frontend (React/Vue/Next.js): Dashboard for businesses + embeddable chat widget
• Backend (Golang):
  • Auth and User/Bot management
  • File ingestion and document parsing
  • Embedding and chunking logic
  • Vector DB integration
  • Prompt building and LLM integration
• Vector Database (e.g., Qdrant, Pinecone):
  • Stores embeddings per bot (namespaced)
• LLM Provider (e.g., OpenAI, Claude):
  • Generates final response using top-matching vectors

---

🧭 Message Flow

1. User submits query to /api/chat
2. Backend:
   • Fetches bot config by bot_id
   • Looks up vector namespace from DB (e.g., vec_abc123)
   • Embeds user message
   • Queries vector DB for top-k similar chunks
   • Builds prompt with retrieved context
   • Calls LLM API and returns response

---

🗂️ Example DB Schema

bots table

Field	Type	Description
id	UUID	Primary key
user_id	UUID	Bot owner
name	VARCHAR	Bot name
vector_namespace	VARCHAR	Vector DB namespace
settings	JSON	Greeting, tone, etc.
created_at	TIMESTAMP	Creation date

files table

Field	Type	Description
id	UUID	File ID
bot_id	UUID	Linked bot
filename	VARCHAR	Name of the uploaded file
type	VARCHAR	pdf, docx, text, url
status	VARCHAR	indexed, failed, pending
uploaded_at	TIMESTAMP	Upload time

---

⚙️ Vector Store Setup

• Each bot gets a unique namespace in the vector DB
• Namespace is stored in the bots table under vector_namespace
• Example: vec_abc123 used for all embeddings of that bot

---

🛠️ Tech Stack

• Backend: Golang (Gin/Fiber/Echo)
• Vector Store: Qdrant / Pinecone
• Embeddings: OpenAI Embedding API (text-embedding-3-small)
• LLM: OpenAI GPT-4 or Claude
• Frontend: React / Next.js
• Auth: JWT-based (email/password or OAuth)

---

⚙️ Configuration

The application supports two configuration methods:

1. Environment Variables (Recommended)

Create a .env file in the project root:

# Copy the example file
cp .env.example .env

# Or migrate from existing config.json
./migrate-config.sh

Edit .env with your actual values. See ENV_CONFIG.md for detailed documentation.

2. config.json (Fallback)

If environment variables are not set, the application falls back to ./app/config/config.json.

Priority: Environment variables take precedence over config.json.

For detailed configuration options, migration guide, and security best practices, see ENV_CONFIG.md.

---

🔐 Authentication

• Use JWT for all authenticated routes
• Authorization header required for:
  • /api/upload
  • /api/chat
  • /api/bots
  • /api/files

---

🧪 Testing the Bot

Once a bot is created:

• Upload a PDF via /api/upload
• Query the bot via /api/chat with:

{
  "botId": "abc123",
  "message": "What services do you offer?"
}