🌐 Universal Offline AI Chatbot

Build your own domain-specific chatbot — offline, modular, and blazing fast.

The Universal Offline AI Chatbot is a privacy-respecting, offline-ready assistant that can chat over any set of PDFs. It’s ideal for legal, cybersecurity, academic, enterprise, or technical domains.

It uses a locally hosted LLM (mistral:instruct via Ollama) and semantic search powered by HuggingFace embeddings and FAISS. You get fast, accurate responses, without sending anything to the cloud.

GitHub Link: https://github.com/AdityaBhatt3010/Universal-Offline-AI-Chatbot
DockerHub Link: https://hub.docker.com/r/adityabhatt3010/ai-chatbot

✨ Highlights

• 🔐 Fully offline-capable with local LLM (via Ollama)
• 📄 Works out-of-the-box with your PDFs
• 🧠 Semantic vector search using all-MiniLM-L6-v2
• ⚡️ Fast and responsive using FAISS backend
• 🧩 Modular, extendable architecture (Streamlit frontend + CLI)
• 🐳 Docker-ready for deployment
• 📸 UI Preview with screenshots
• ✅ Built-in CI/CD check via GitHub Actions
• 🎯 Fully reproducible setup via PowerShell script or Docker

🧱 Tech Stack

⚠️ HuggingFace Token is required to fetch the embedding model once. It's cached locally afterward.

Example .env:

HF_TOKEN=your_huggingface_token_here

💡 Use Cases

📁 Project Structure

Universal-Offline-AI-Chatbot/
│
├── data/                   # Place your PDF documents here
│   └── Try.pdf
│
├── Screenshots/           # UI snapshots
│   ├── Loading_Screen.png
│   └── Running_the_Model.png
│
├── src/                   # Modular source code
│   ├── chunker.py
│   ├── config.py
│   ├── embedding.py
│   ├── loader.py
│   ├── model_loader.py
│   ├── prompts.py
│   ├── qa_chain.py
│   ├── utils.py
│   └── vectorstore.py
│
├── vectorstore/           # Local FAISS vector index
│   └── db_faiss/
│
├── Bot.py                 # CLI script
├── Bot.ipynb              # Jupyter notebook version
├── main.py                # Entry-point (optional)
├── streamlit_app.py       # Frontend UI (Streamlit)
├── requirements.txt       # Python dependencies
├── setup.ps1              # PowerShell setup script
├── Dockerfile             # Docker image definition
├── .dockerignore
├── .env                   # Contains HF_TOKEN
├── README.md
└── LICENSE

🧰 Setup Instructions

🖥️ One-Click Setup (Windows Only)

.\setup.ps1

This will:

• Create virtual env
• Install dependencies
• Pull Mistral via Ollama
• Ask for Hugging Face token
• Build Docker image

🛠 Manual Setup

1. Install Python Requirements

pip install -r requirements.txt

2. Install & Pull Ollama Model

ollama pull mistral:instruct

3. Set HuggingFace Token (First Time Only)

export HUGGINGFACEHUB_API_TOKEN=your_token      # macOS/Linux
set HUGGINGFACEHUB_API_TOKEN=your_token         # Windows CMD

4. Run the CLI Bot

python Bot.py

🌐 Run with Streamlit Frontend

streamlit run streamlit_app.py

📸 Streamlit Preview

⏳ Loading Screen

🤖 Chat in Action

🐳 Docker Support

Prerequisites

• Docker installed & running
• .env file containing HF_TOKEN (Hugging Face token)

🛠️ Docker Build & Run

To build and run the chatbot using Docker, follow these steps:

1. Build the Docker image:

   docker build -t ai-chatbot .
   

2. Run the container (with volume mount and token):

   docker run -p 8501:8501 --env-file .env -v ${PWD}/data:/app/data ai-chatbot
   

   This will:

   • Map the container's port 8501 to local 8501
   • Use your local .env for HF_TOKEN
   • Mount the data/ folder into the container for access to PDFs

Access the chatbot at http://localhost:8501

📸 Screenshot:

🐳 Building the Image

🔄 Using Your Own PDFs

# Replace default file(s)
mv your_files/*.pdf ./data/

# Re-run the bot or restart Streamlit
python Bot.py

Automatically re-indexes your new documents using FAISS.

🧪 Sample Interaction

🧠 You: What does Article 21 state?

🤖 Bot: Article 21 of the Indian Constitution guarantees the protection of life and personal liberty...

🧑‍💻 Author

Aditya Bhatt
Cybersecurity Specialist | VAPT Expert | OSS Contributor
GitHub | Medium