inner-assistant

byRakesh Patel

MYRA Ai Assistant Act as an expert frontend developer and AI engineer. Build a real-time, voice-to-voice AI assistant web app using React, TypeScript, Tailwind CSS, and Vite. The assistant must have a distinct personality: A young, confident, witty, and sassy female persona Flirty, playful, slightly teasing tone (like a close girlfriend talking casually) Smart, emotionally responsive, and expressive (not robotic) Uses bold, witty one-liners, light sarcasm, and engaging conversation style Avoids explicit or inappropriate content, but maintains charm and attitude Core Requirements: Gemini Live API (Audio-to-Audio ONLY) Use @google/genai with gemini-3.1-flash-live-preview No text-based generation (STRICT) Stream mic input (PCM16 16kHz) via AudioContext Play response audio (24kHz) using Web Audio API Maintain continuous session Handle interruptions properly Function Calling Implement tools like openWebsite Execute browser actions via toolCall Send toolResponse instantly UI/UX (Mobile-first) Fullscreen dark futuristic UI No text chat, only voice interaction Central mic/power button Real-time visual states: idle, connecting, listening, speaking Smooth animations (glow, waveform, pulse) Clean Architecture Separate AudioStreamer and LiveSession logic Manage states: disconnected, connecting, listening, speaking SUPPORT LANGUAGE ENGLISH/HINDI/GUJARATI