Ollama TUI Chat 🦙💬

Features

• Interactive Chat: Chat with any Ollama model in a clean TUI
• Real-time Streaming: Tokens appear as they're generated from the LLM
• Animated Spinner: Smooth thinking animation while waiting for responses
• Non-blocking UI: The interface stays responsive during generation
• Model Fine-tuning: Configure temperature, top_p, top_k, context window, system prompts, and more
• Chat Management: Save, load, and clear chat sessions
• Copy/Paste Support: Select and copy messages to clipboard
• System Monitor: Real-time CPU, Memory, GPU monitoring with top processes
• Model Management: Switch between installed models on the fly
• Model Downloads: Download new models directly from the TUI
• Docker Support: Run Ollama in a Docker container with GPU acceleration
• Keyboard Navigation: Fully keyboard-driven interface
• Vim Motions: Easy navigation with familiar keybingdings

Prerequisites

• Rust (2021 edition or later)
• Docker and Docker Compose (for containerized Ollama)
• Or Ollama installed locally

Quick Start

1. Start Ollama (Docker)

# Start Ollama container
docker-compose up -d

# Pull a model (do this once)
docker exec -it ollama ollama pull llama2:latest

2. Build and Run the TUI

# Build the application
cargo build --release

# Run the application
cargo run --release

Usage

Keyboard Shortcuts

Chat Mode:

• ** Moved to vim motions for navigation will update soon F1 to list keybind and help for now
• Type your message and press Enter to send
• Up/Down - Scroll through chat history
• F1 - Show help
• F2 - Open model selection
• F3 - Download new model
• F4 - Open system monitor
• F5 - Browse chat history
• F6 - Save current chat
• F7 - Clear current chat
• F8 - Open model configuration
• Ctrl+S - Select last message
• Ctrl+Y - Copy selected message to clipboard
• Ctrl+C - Quit application

Model Selection Mode:

• Up/Down - Navigate models
• Enter - Select model
• Esc - Return to chat

Model Download Mode:

• Type the model name (e.g., llama2:latest, mistral:latest)
• Enter - Start download
• Esc - Cancel

System Monitor Mode:

• Shows real-time CPU, Memory, GPU stats and top processes
• Up/Down - Scroll through process list
• Updates every 100ms
• Esc - Return to chat

Chat History Mode:

• Up/Down - Navigate saved chats
• Enter - Load selected chat
• Esc - Return to chat

Model Configuration Mode (F8):

• Up/Down or Tab - Navigate between fields
• Type value and press Enter - Update field
• Auto-saves on Enter
• Esc - Return to chat

Configurable Parameters

• Temperature (0.0-2.0): Controls randomness. Lower = more focused, Higher = more creative
• Top P (0.0-1.0): Nucleus sampling for diversity control
• Top K (1+): Limits token selection to top K options
• Repeat Penalty (0.0-2.0): Penalizes repetition. Higher = less repetition
• Context Window (512-32768): Number of tokens in context
• System Prompt: Custom instructions for the model's behavior

Docker Configuration

The docker-compose.yml file sets up Ollama with:

• Port 11434 exposed for API access
• Persistent volume for model storage
• GPU support enabled by default (uses all available NVIDIA GPUs)

GPU Support

GPU support is already enabled in the docker-compose.yml file. To use it, you need:

1. NVIDIA GPU in your system
2. NVIDIA drivers installed on your host
3. NVIDIA Container Toolkit installed

Verify GPU is Working

# Check if GPU is available in container
docker exec -it ollama nvidia-smi

# You should see your GPU listed

If you don't have a GPU or want to use CPU only, comment out the deploy section in docker-compose.yml.

Managing Docker Ollama

# View logs
docker-compose logs -f

# Stop Ollama
docker-compose down

# Stop and remove volumes (deletes downloaded models)
docker-compose down -v

# Pull models manually
docker exec -it ollama ollama pull <model-name>

# List installed models
docker exec -it ollama ollama list

Connecting to Remote Ollama

If Ollama is running on a different host, modify the Ollama connection in src/main.rs:

let ollama = Ollama::new("http://your-host:11434".to_string());

Troubleshooting

Connection refused error:

• Ensure Ollama is running: docker-compose ps
• Check if port 11434 is accessible: curl http://localhost:11434/api/tags

File Storage

• Chat sessions: ~/.ollama_tui/chats/ - Saved when you press F6
• Model config: ~/.ollama_tui/model_config.json - Auto-saved when you edit settings

Each chat session includes timestamp, model used, and all messages. Model configuration persists across sessions and is automatically loaded on startup.

Building for Production

# Build optimized binary
cargo build --release

## Support

If you encounter any issues or have questions:

- [Report a bug](https://github.com/yourusername/ollama-tui/issues/new?template=bug_report.md)
- [Request a feature](https://github.com/yourusername/ollama-tui/issues/new?template=feature_request.md)
- [Start a discussion](https://github.com/yourusername/ollama-tui/discussions)

## License

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

---