vision

Analyze images, audio, video, or PDFs using Google Gemini via Vertex AI (GCP credits). Generic multimodal router. Use when the user explicitly says "analyze with gemini", "use gemini vision", "run through gemini", "transcribe with gemini", "describe with gemini", or invokes /vision. Works on any media file the caller supplies — photos, screenshots, diagrams, UI, products, art, audio, video, documents. NOT for: live streams, real-time analysis, or when the host model's own vision/audio is acceptable.