Voice memory wip

[Pash10g]

This pull request introduces the initial implementation of the Voice Memory Demo app, a Next.js project that showcases persistent memory for voice AI agents using Gemini Live and MongoDB. The changes include core backend and frontend setup, environment and configuration files, documentation, and essential utilities for audio processing and memory management.

The most important changes are:

Core Application Features and API Endpoints

• Implements the main memory API endpoint in route.ts for CRUD and search operations on user memories, supporting operations like set, get, delete, query, and list.
• Adds the Gemini session configuration endpoint in route.ts, providing system instructions and tool definitions for the voice agent, including persistent memory capabilities.

Frontend and UI

• Sets up the main application entry point and layout, including a configuration check and integration of the VoiceAgent component for the interactive voice experience. [1] [2]
• Adds global styles and a dark theme, including custom scrollbar and recording indicator animation.

Configuration, Documentation, and Utilities

• Provides a comprehensive README.md with setup instructions, architecture, features, and troubleshooting for the demo app.
• Adds environment variable templates, .gitignore, and Next.js/Tailwind/PostCSS/ESLint config files for project setup and best practices. [1] [2] [3] [4] [5]
• Implements an AudioWorklet processor in audio-processor.js for capturing and converting microphone audio to the required PCM format for Gemini Live API.
• Adds package.json with all necessary dependencies for Next.js, MongoDB, Gemini API, Tailwind CSS, and development tooling.