🚀 Next-Gen Private AI Chat - Your Data, Your Control

Experience the future of confidential conversations with self-hosted AI models running entirely on your device

A private, free, offline-first chat application powered by Open Source AI models like DeepSeek, Llama, Mistral, etc. through Ollama.

Features

• Full conversation history
• Image upload support for vision models (LLaVA, Llama3.2-Vision)
• Model parameter controls
• Ollama-powered local AI processing
• Multi-LLM support (DeepSeek, Llama 2, Mistral, etc.)
• Markdown support
• System prompt engineering
• Secure offline storage
• Open source model ecosystem

Tech Stack

Frontend: HTML5, CSS3, Vanilla JavaScript
Backend: FastAPI, Ollama Python Client
AI Engine: Ollama (Free & Open Source)

• Local model deployment & management
• Cross-platform support (Linux/macOS/Windows)
• Community-driven model library (Llama 2, Mistral, etc.)
• Efficient CPU/GPU resource utilization
  Powered By: DeepSeek-R1

🛠️ Tech Powerhouse

graph LR
A[Beautiful UI] --> B[Vanilla JS]
B --> C{FastAPI}
C --> D[(Ollama Engine)]
D --> E[Local Models]
E --> F[CPU/GPU Optimized]

Loading

🌐 Supported Model Ecosystem

Model Family	Example Sizes	Specialization
Vision Models	7B, 13B, 34B	Image understanding
DeepSeek	7B, 33B	General purpose, coding
Llama 3	8B, 70B	General reasoning
LLaVA	7B, 13B	Multimodal vision
Llama3.2-Vision	8B	Image-to-text

❗ Important Note: Image upload functionality only works with vision-capable models like:

• llava:7b - Large Language and Vision Assistant
• llama3.2-vision - Llama3-based visual understanding
• llava-llama3 - LLaVA implementation for Llama3

❗ System Requirements Note
⚠️ Memory Requirements for Local AI:

• 7B models: Minimum 8GB RAM (16GB recommended)
• 13B models: Minimum 16GB RAM (32GB recommended)
• 33B+ models: 32GB+ RAM required
  Note: Insufficient memory may cause slow performance or model failures

Installation

1. Ollama Setup

Follow platform-specific instructions:

macOS

curl -fsSL https://ollama.com/install.sh | sh

or

Download installer from ollama.com/download/Ollama-darwin.zip

Windows

Download installer from ollama.com/download/OllamaSetup.exe

Linux

curl -fsSL https://ollama.ai/install.sh | sh

2. Backend Setup

2.1. Start Ollama Service:

ollama serve

# Optional: Pull models - for faster startup. Note You can also download models from chat.html
ollama pull deepseek-r1  # 33B parameter model
ollama pull llama2:13b   # Medium-sized variant
ollama pull llava:7b        # 7B parameter vision model
ollama pull llama3.2-vision    # Official Llama3 vision variant
ollama pull llava-llama3:8b    # Llama3-based LLaVA implementation

2.2. Install Python Dependencies:

pip install -r backend/requirements.txt

2.3. Start FastAPI Server:

cd backend
uvicorn main:app --reload --port 3000

3. Frontend Setup (Required for Offline Use)

# Navigate to frontend directory
cd frontend

# Serve via local HTTP server on port 8000
python3 -m http.server 8000 --bind 127.0.0.1

# Now access at: http://localhost:8000/chat.html

localhost:8000/chat.html

4. Alternative Access Methods

Method	Command	When to Use
Local Server (Recommended)	python3 -m http.server 8000	Offline access
Direct File	open chat.html	(Requires Internet Access in case of file access)

❗ Critical Offline Note:
For complete offline functionality:

1. First download models while online using the interface
2. All services must run on localhost:
   • Ollama: ollama serve
   • Backend: uvicorn main:app --port 3000
   • Frontend: python3 -m http.server 8000
3. Access via http://localhost:8000/chat.html never file://

Running Full Stack

sequenceDiagram
    participant User
    participant Frontend
    participant Backend
    participant Ollama
    
    User->>Frontend: http://localhost:8000
    Frontend->>Backend: API Calls (localhost:3000)
    Backend->>Ollama: Model Requests (localhost:11434)
    Ollama-->>Backend: AI Responses
    Backend-->>Frontend: Processed Data
    Frontend-->>User: Rendered Interface

Loading

Project Structure

project-root/
├── frontend/
│   ├── chat.html        # Main interface
│   ├── app.js              # Client-side logic
│   └── styles.css          # Styling
└── backend/
    ├── main.py             # FastAPI server
    └── requirements.txt   # Python dependencies

---

📜 Full Documentation
Explore our API Doc for:

• Advanced model configurations
• Vision pipeline architecture
• Local deployment guides
• Plugin development tutorials

Image Support Details

• 📸 Multi-image conversations (JPG/PNG/GIF)
• 🔍 Zoomable previews with pinch-to-zoom
• 🧠 Vision model integration
• 📦 Local image processing (no cloud uploads)

⚠️ Important: Image understanding requires vision-capable models like: llava:7b, llama3.2-vision, llava-llama3:8b

Capabilities:

• Upload multiple images per message (JPG/PNG)
• Images are processed locally using vision models
• Combined text+image understanding
• Image previews with zoom functionality

Requirements:

1. Must use a vision-capable model
2. Images under 5MB recommended
3. Supported formats: JPEG, PNG, GIF

Disclaimer:
⚠️ Standard language models (non-vision) will ignore attached images.
🔍 Vision capabilities require specific model architectures - regular LLMs cannot process images even if attached.

Acknowledgements

• Special thanks to MK.
• Ollama integration (Apache 2.0 License) for private AI processing
• UI design influenced by modern chat applications

💡 Ethical AI Note

We champion responsible AI - all models are open-source and locally hosted.
No hidden trackers, no data harvesting, just pure AI innovation.

License

AGPL-3.0

---

🔗 Powered By
Ollama • DeepSeek-R1 • FastAPI

Port Configuration Reference

Service	Port	Purpose
Frontend	8000	Web interface
Backend	3000	FastAPI server
Ollama	11434	Model inference & management