Antigravity Vision Proxy

Overview

Use agy --print as a subprocess to analyze images via Gemini vision models. Frame extraction (ffmpeg) + agy invocation + structured output parsing.

When to Use

• Agent lacks built-in vision but needs to identify game, heroes, or activity from video frames
• Video frame analysis needed for timestamping or scene classification
• Any scenario requiring visual context that the current model cannot see

Modality Comparison: Text-Only vs. Text + Vision