🤖 Domain-Chatbot

A Local Domain-Based AI Chatbot

Developer: Sabithulla

---

📌 Overview

Domain-Chatbot is a fully local, domain-aware AI chatbot that changes its behavior based on the selected domain:

• 💬 Chat → Friendly, casual conversations
• 💻 Coding → Programming help with syntax highlighting
• 📚 Exam → Exam-oriented answers with structured points and formulas

The project runs 100% offline, uses Ollama for local LLMs, and supports real-time streaming, KaTeX math rendering, and code highlighting — all without any API keys.

---

✨ Why Domain-Chatbot?

Most chatbots treat every question the same.

Domain-Chatbot doesn’t.

• 💬 Chat mode → friendly, human-like conversation
• 💻 Coding mode → clean code + syntax highlighting
• 📘 Exam mode → structured answers + KaTeX formulas

All of this runs locally — no API keys, no internet dependency.

---

🖼️ Screenshots

💬 Chat Domain

💻 Coding Domain

📘 Exam Domain (KaTeX formulas)

---

🧪 Example Prompts

Chat

Explain black holes like I'm 10 years old

Coding

Write a Python program to check whether a number is palindrome

Exam

Derive Ohm’s law and explain with formula

---

✨ Key Features

• 🔁 Real-time streaming responses
• 🧠 Domain-based AI behavior
• 🧮 Automatic KaTeX rendering for formulas
• 🖍️ Syntax highlighting for code
• 💡 Centered suggestion prompts (one-time per domain)
• 🔒 Fully offline & privacy-friendly
• ⚡ Lightweight frontend (no frameworks)

---

🧠 Domain Behavior

Domain	Behavior	Special Handling
Chat	Friendly & conversational	Natural responses
Coding	Developer-focused	Highlight.js for code
Exam	Exam-oriented	KaTeX for formulas

---

🏗️ Architecture

Frontend (HTML + CSS + JavaScript)
↓
FastAPI Backend
↓
Ollama Local Models

---

🧮 Smart Formula Detection (KaTeX Logic)

The chatbot intelligently decides what should be rendered as math.

Rules:

• Text inside ** ** containing math symbols (= + - * / ^) → rendered using KaTeX
• Text inside ** ** without math symbols → rendered as bold text
• Normal text remains unchanged

Examples:

• R = V / I → KaTeX formula
• Resistance → Bold text
• Explanation: → Normal text

This avoids converting headings or explanations into formulas.

---

🛠️ Tech Stack

Backend

• Python 3.9+
• FastAPI
• Uvicorn
• Requests
• Ollama (Local LLM runtime)

Frontend

• HTML
• CSS
• JavaScript
• KaTeX (Math rendering)
• Highlight.js (Code highlighting)

AI Models (via Ollama)

• qwen2.5:0.5b → Chat
• qwen2.5-coder:0.5b → Coding
• gemma:2b → Exam

---

📂 Project Structure

Domain-Chatbot
├── backend
│ ├── main.py
│ ├── router.py
├── frontend
│ ├── index.html
├── assets
│ ├── chat.png
│ ├── coding.png
│ ├── exam.png
├── README.md

---

⚙️ Installation & Setup

Requirements

• Python 3.9 or above
• Ollama installed locally

Ollama Models Required

• qwen2.5:0.5b
• qwen2.5-coder:0.5b
• gemma:2b

Python Dependencies

• fastapi
• uvicorn
• requests

---

▶️ How to Run

1. Start Ollama locally
2. Run FastAPI backend
3. Open frontend/index.html in browser
4. Select a domain
5. Use suggestion prompts or type your own query
6. Watch responses stream live

---

👨‍💻 Developer

Sabithulla
Built with curiosity, logic, and a focus on clean AI user experience.

---

⭐ Support

If you found this project useful:

• ⭐ Star the repository
• 🍴 Fork it
• 🧠 Extend it with new domains

---

🔮 Roadmap

• Voice input and output
• Mobile-friendly UI
• More domains (Medical, Legal, Interview)
• Model selector in UI

---

📜 License

This project can be released under the MIT License. You are free to use, modify, and distribute it.