aidash — AI Coding Assistant Token & Cost Dashboard

A rich terminal dashboard for tracking token usage, costs, and session metrics across GitHub Copilot CLI and Claude Code.

Built in Rust with ratatui — line charts, bar charts, live monitoring, and multi-view navigation.

Features

Dashboard

• Sessions table — All sessions at a glance with status, name, model, tokens, cost, turns, agents, duration
• Multi-column sorting — Press s to cycle, S to reverse, or 1–9 to jump to a specific column
• Source filtering — Toggle between All / Copilot / Claude with Tab
• Active sessions first — Default sort groups active (● ON) sessions at the top

Session Detail (Enter)

• Token breakdown — Output, estimated input, reasoning tokens, total, and cost
• Output vs Input ratio bar — Visual split of token distribution
• Per-turn token chart — Line chart (Chart widget with Braille markers) showing token usage over turns
• Models used — List of models with message counts, colored by tier
• Sub-agents summary — Grouped by type with total duration
• Top tools — Mini bar chart of the 10 most-used tools
• Code changes — Lines added/removed, files modified

Live Dashboard (L / l)

• 4-quadrant real-time view for active sessions, auto-refreshing every 2 seconds
• Token accumulation — Dual-line chart: per-turn tokens + cumulative growth curve
• Tool activity — Horizontal bar chart of tool call frequency
• Model distribution — Bar chart of messages per model
• Sub-agents panel — Grouped stats + currently running agents with elapsed time
• Rate metrics — Tokens/min, live cost ticker, turn count

Additional Views

• Agents (a) — Detailed table of all sub-agent tasks with type, timing, duration, status
• Tools (t) — Full bar chart with gradient colors, percentages, category-colored names
• Models (m) — Per-model message distribution, cost attribution, and details table
• Trends (T) — 30-day line chart (tokens + cost overlay), model distribution, top tools, session timeline
• Help (?) — Full keybinding reference overlay

Themes

• Dark theme (default) — Warm gold/amber/lime palette on dark background
• Light theme (--light flag or d key) — Deep tones on warm white background
• Toggle at runtime with d key

CLI Commands

aidash                    # Launch interactive TUI
aidash list               # Print sessions table
aidash list --json        # JSON output for scripting
aidash session <id>       # Session details (non-interactive)
aidash cost               # Total cost summary
aidash cost --since DATE  # Cost since YYYY-MM-DD
aidash init               # Create ~/.aidash/pricing.json
aidash update-pricing     # Refresh pricing with latest defaults
aidash --source copilot   # Filter to Copilot sessions only
aidash --source claude    # Filter to Claude sessions only
aidash --light            # Start with light theme

Installation

From source

git clone https://github.com/user/aidash.git
cd aidash
cargo build --release
# Binary at target/release/aidash

Requirements

• Rust 1.70+ (install via rustup)
• No other dependencies — SQLite is bundled

Keyboard Shortcuts

Dashboard

Key	Action
↑/↓ or j/k	Navigate sessions
Enter	Drill into selected session
s	Cycle sort column
S	Reverse sort direction
1–9	Sort by column N directly (toggles direction)
Tab	Switch source: All → Copilot → Claude
L	Live dashboard (active sessions only)
T	Trends & analytics
r	Refresh data from disk
d	Toggle dark/light theme
?	Help overlay
q	Quit

Session Detail

Key	Action
a	Sub-agents breakdown
t	Tool usage breakdown
m	Models breakdown
l	Live dashboard (active sessions)
Esc	Back to dashboard

All Views

Key	Action
d	Toggle theme
Esc / Backspace	Go back
q	Quit

Data Sources

GitHub Copilot CLI (~/.copilot/)

File	Data
session-state/<id>/events.jsonl	Token counts, sub-agents, tool calls, model changes
session-state/<id>/workspace.yaml	Session metadata (summary, branch, cwd)
session-store.db	SQLite — session summaries, turns
config.json	Current model selection

Claude Code (~/.claude/)

File	Data
projects/<project>/<session>.jsonl	Conversation history, token usage, cost
history.jsonl	Global session index

Custom Pricing

Pricing is stored in ~/.aidash/pricing.json (auto-created on first run).

Edit it to add custom models or override prices:

{
  "claude-opus-4.6-1m": {
    "input_per_million": 15.0,
    "output_per_million": 75.0,
    "is_premium": true
  },
  "my-custom-model": {
    "input_per_million": 1.0,
    "output_per_million": 5.0,
    "is_premium": false
  }
}

Run aidash update-pricing to merge the latest default prices into your config without losing custom entries.

Architecture

src/
├── main.rs              # CLI entry point (clap)
├── app.rs               # TUI state machine, event loop, view routing
├── config.rs            # Pricing configuration (~/.aidash/pricing.json)
├── util.rs              # Number/time/string formatting helpers
├── data/
│   ├── models.rs        # Unified Session, Turn, SubAgent, ToolCall types
│   ├── copilot.rs       # Copilot CLI event parser (events.jsonl + SQLite)
│   └── claude.rs        # Claude Code session parser (JSONL)
├── cost/
│   └── estimator.rs     # Token estimation & cost calculation
└── tui/
    ├── theme.rs         # Dark/light theme system with warm color palette
    ├── dashboard.rs     # Sessions overview table with sorting
    ├── session.rs       # Session detail (charts, token breakdown)
    ├── live.rs          # Real-time monitoring dashboard (4 quadrants)
    ├── agents.rs        # Sub-agents detail table
    ├── tools.rs         # Tool usage bar chart
    ├── models.rs        # Per-model analysis with cost attribution
    ├── trends.rs        # 30-day trends, sparklines, timeline
    └── help.rs          # Help overlay

Token Estimation

Since exact input tokens aren't tracked in Copilot events, aidash estimates:

• Output tokens — Directly from assistant.message.outputTokens
• Input tokens — Estimated from user message content length (~4 chars/token)
• Reasoning tokens — Estimated from reasoningText length
• Cost — Calculated per-model using the pricing table, distributed proportionally by message count

Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Run cargo build and cargo clippy to verify
5. Submit a pull request

License

MIT