User Guide & Dashboard

The Antigravity Usage Intelligence interface consists of two synchronized surfaces:

The Left/Right Sidebar View (Agent Stats container in the Activity Bar) for quick glance telemetry while coding.
The Full-Screen Dashboard (opened via Ctrl+Shift+P → Antigravity Stats: Open Full Screen Dashboard or clicking the expand icon).

Both views share the same state and automatically adapt to both Dark Theme and Light Theme in real time.

⚡ 1. Model Quota & Subscription Status

The Model Quota card gives real-time visibility into your Google Antigravity quotas.

Subscription Tier Badge: Automatically detects if you are on AI Pro, Ultra, or free tier.
Live Reset Countdown: A real-time countdown timer (e.g. 3h 18m) ticking down to when your quota pool refreshes.
Model Family Progress Bars:
- Gemini 3.8 Flash & Gemini 3.7 Flash
- Claude Sonnet 4.6 & Claude Opus 4.6 (Thinking)
Threshold Alert Indicators: Turns from subtle green to amber (75%), deep orange (90%), or red (100%) when quota approaches exhaustion.

The Heatmap maps your AI agent pairing activity across the calendar year.

GitHub Canonical Grid: 7-day weekday matrix from Monday to Sunday with Mon, Wed, Fri labels and month markers.
4-Level Emerald Intensity Scale:
- Level 0: GitHub neutral grey (Light mode #ebedf0, Dark mode #161b22)
- Level 1: Light activity (1 to 20K tokens)
- Level 2: Moderate pairing (20K to 100K tokens)
- Level 3: Deep development (100K to 500K tokens)
- Level 4: Heavy agent workloads (> 500K tokens)
Streak & Peak Statistics:
- Active Days: Total days with active AI pairing.
- Consistency %: Percentage of days with recorded turns.
- Peak Day: Your highest single-day token consumption date and count.
- Current & Best Streaks: Daily consecutive streak tracking.
1-Click Historical Filtering: Click any tile to isolate the dashboard to that specific date. Click Filter Dashboard to All History to return to the full view.

Identifies when you and your autonomous agents produce the highest volume of work throughout the 24 hours of the day.

Hourly Bar Distribution: Aggregates token volume and session starts into hourly bins (12a, 3a, 6a, 9a, 12p, 3p, 6p, 9p).
Golden Peak Highlight: Automatically isolates your most productive hour (e.g. 🔥 Peak: 3:00 PM (34 sessions)).

Antigravity Usage Intelligence categorizes tokens into four distinct architectural streams:

🟦 Fresh Input Tokens: Raw prompt tokens sent to the model for new or changed context.
🟪 Prompt Cache Hits: Context tokens served directly from Gemini's prompt cache at ultra-low latency and zero additional cost.
🟩 Generation Output Tokens: Code completions, diff chunks, and final agent responses.
🟧 Thinking / Reasoning Tokens: Internal chain-of-thought tokens produced by hybrid thinking models (Gemini Flash Thinking / Claude Opus Thinking).

Clicking on any individual session in the recent session list opens a slide-out drawer providing forensic audit logs:

Session ID: Full UUID with 1-click clipboard copy.
Workspace Path: Root directory where the agent turn occurred.
Duration & Turn Count: Total conversation runtime and turns.
Tool Invocation Breakdown: Color-coded badges tracking agent commands:
- run_command (Shell execution)
- replace_file_content (Precise code edits)
- view_file (Context reading)
- grep_search (Codebase indexing)
- write_to_file (File creation)
Markdown Export: 1-click button generates a structured markdown summary ready to paste into PR descriptions or issue reports.

Antigravity Usage Intelligence · 100% Offline & Private · Open Source under MIT License · GitHub