Summary
- Client-side Gemini integration for web apps with multimodal inference, streaming, and on-device hybrid execution.
- Supports text-only and multimodal inputs (images, audio, video, PDFs); files over 20 MB route through Cloud Storage Includes chat sessions with automatic history, streaming responses for real-time display, and structured JSON output enforcement Offers hybrid on-device inference via Gemini Nano in Chrome, with automatic fallback to cloud execution Requires App Check for production security and supports dynamic model versioning through Remote Config to avoid redeployment Choose between Gemini Developer API (free tier, prototyping) or Vertex AI Gemini API (enterprise scale, requires Blaze plan)