tui-agent-dashboard

A beautiful terminal UI dashboard for monitoring AI agent operations, costs, and attribution in real-time—think htop meets agent observability.

What is this?

tui-agent-dashboard brings observability to your AI agent infrastructure with a gorgeous, keyboard-driven terminal interface. It integrates seamlessly with your existing agent-cost-attribution-layer and ai-agent-spend-guardian systems to surface real-time metrics, cost breakdowns, and request attribution without leaving your terminal. Built for developers who prefer TUIs over web dashboards.

Features

• Live Agent Activity Panel – Monitor active agents, request throughput, and execution status in real-time
• Cost Attribution Dashboard – View API costs broken down by agent, model, and time period
• Request Inspector – Search and inspect individual requests with full attribution metadata
• Health & Performance Metrics – Track agent health, latency, and error rates at a glance
• Interactive Navigation – Vim-style keybindings (j/k scroll, tab to switch panels, / search)
• Export Capabilities – Dump data to JSON/CSV or take ASCII screenshots for sharing
• Multiple Themes – Dark and light modes with configurable refresh rates
• Plugin Support – Out-of-the-box integration with LiteLLM, LangChain, and raw OpenAI/Anthropic proxies
• Docker Ready – Run locally with docker-compose or deploy anywhere

Quick Start

Installation

pip install tui-agent-dashboard

Configuration

Copy .env.example to .env and configure your data sources:

cp .env.example .env

Update with your agent infrastructure endpoints:

AGENT_ATTRIBUTION_API_URL=http://localhost:8000
SPEND_GUARDIAN_API_URL=http://localhost:8001
REFRESH_RATE_MS=1000

Run

tui-agent-dashboard

Or with Docker:

docker-compose up

Usage

Once running, interact with the dashboard using keyboard commands:

Example: Monitor costs by agent

1. Launch the dashboard
2. Press Tab until you reach the Costs panel
3. Use j/k to browse agents sorted by spend
4. Press e to export to CSV

Example: Export metrics

Press e from any panel to export real-time data:

# Exported to:
# tui-agent-dashboard-2025-05-04T09-03-42.json
# tui-agent-dashboard-2025-05-04T09-03-42.csv

Tech Stack

• Framework: Textual – Modern Python TUI library with GPU-accelerated rendering
• Backend Integration: FastAPI-compatible REST client
• Data Models: Pydantic for type-safe data validation
• Testing: pytest with comprehensive integration tests
• Containerization: Docker & Docker Compose
• Python: 3.9+

Development

Clone and install in editable mode:

git clone <repo>
cd tui-agent-dashboard
pip install -e ".[dev]"

Run tests:

pytest tests/

See CONTRIBUTING.md for contribution guidelines.

License

MIT