Trackio VibeTUI

⚠️ Notice: This project is vibe-coded and not thoroughly tested or even fully implemented. Users may be interested in or prefer to use either trackio-view (Python) or trackio-tui (Rust).

Terminal User Interface for Trackio ML experiment tracking. View your training metrics, system statistics, and experiment runs directly in your terminal.

Features

• 📊 Interactive metric visualization with line plots
• 🎨 Color-coded multi-run comparison
• 🔍 Metric filtering and search
• ⚙️ Configurable chart controls (smoothing, log scales, x-axis)
• 🖥️ Clean terminal interface using Textual framework
• ⚡ Async data loading with caching

Installation

pip install -e .

Or with development dependencies:

pip install -e ".[dev]"

Usage

Launch the TUI:

trackio-vibetui

Or run as a Python module:

python -m trackio_vibetui

Keyboard Shortcuts

• q - Quit application
• r - Refresh data (clear cache)
• m - Show metrics view
• s - Show system metrics view (coming soon)
• n - Show runs view (coming soon)
• t - Show media/tables view (coming soon)
• f - Show files view (coming soon)
• ? - Show help

Navigation

• Use mouse or keyboard to interact with widgets
• Tab to move between widgets
• Arrow keys in lists and inputs
• Click buttons in the header to switch between views

Requirements

• Python 3.8+
• Trackio (for ML experiment tracking data)
• Textual (TUI framework)
• textual-plotext (for plotting)

Development Status

✅ Implemented (Phase 1 - MVP)

• Project structure and dependencies
• Data loader with async support and caching
• Metric visualization with PlotextPlot
• Run selector with filtering
• Chart controls (smoothing, log scales, x-axis)
• Main metrics screen
• Basic CSS styling

🚧 Coming Soon (Phase 2+)

• System metrics screen
• Runs table view
• Run details screen
• Media & tables viewer
• Files browser
• Auto-refresh functionality