Glass is a local-first, desktop-based AI assistant built with Electron and Next.js that transforms your screen activity and audio into structured knowledge. Designed for professionals in meetings, research, or content creation, it eliminates the need for manual note-taking by automatically surfacing action items and answers based on your real-time digital context. Unlike cloud-based tools, Glass runs entirely on your machine with no screen recording or hidden data sharing.
Built with Electron, Next.js, Tailwind CSS, and Node.js 20.x, Glass supports OpenAI, Gemini, Claude, and local LLMs via Ollama, with Whisper for speech-to-text. It uses Firebase for data storage and Rust for audio processing, ensuring low-latency, privacy-first operation across macOS, Windows, and Linux.
What You Get
- Real-time Meeting Summaries - Automatically generates meeting notes and action items from live audio and screen activity during calls or presentations.
- Context-Aware AI Queries - Answers questions using your entire history of screen interactions and audio input, not just isolated prompts.
- Local LLM & STT Support - Run models like Ollama and Whisper entirely offline for privacy-sensitive use cases without cloud dependency.
- Multi-Provider AI Integration - Supports OpenAI, Gemini, and Claude APIs with configurable API keys for flexible model selection.
- Invisible Operation - Never appears in screenshots, screen recordings, or system dock; operates without persistent capture or data uploads.
- Cross-Platform Desktop App - Runs natively on macOS, Windows (beta), and Linux using Electron and Rust for audio enhancement and system integration.
Common Use Cases
- Running professional meetings - A project manager uses Glass to auto-generate action items and summaries during Zoom calls without lifting a finger.
- Research and knowledge work - A PhD student captures insights from video lectures and screen-based reading, turning them into structured notes for later review.
- Privacy-conscious content creators - A journalist uses Glass to transcribe interviews and extract quotes without uploading audio to third-party servers.
- Developers documenting workflows - A software engineer uses Glass to log debugging steps and code patterns from their screen, creating a personal knowledge base.
Under The Hood
Architecture
- Clear separation between main and renderer processes using Electron’s IPC system with secure preload scripts
- Modular service layer with abstracted data providers (SQLite, Firestore) and repository patterns for consistent data access
- Decoupled AI provider architecture with request queuing to manage non-streaming API concurrency
- Database abstraction with schema initialization and migration handled by a dedicated initializer
- Deep linking and protocol handling with cross-platform URL sanitization and second-instance session routing
Tech Stack
- Electron-based desktop application with React and Next.js for UI, bundled and signed for native deployment
- Backend services powered by Express and Firebase Admin SDK, with Firestore for real-time sync and Functions for serverless logic
- Multi-cloud AI orchestration via unified API gateway, supporting OpenAI, Anthropic, Google Generative AI, and Deepgram
- Local persistence using SQLite and electron-store, with secure credential storage via keychain integration and image processing via sharp
- Deployment pipeline with automated code signing, platform-specific packaging, and built-in update mechanisms
Code Quality
- Consistent naming conventions and feature-based modular organization enhance maintainability
- Strong type safety and schema constraints ensure data integrity, though error handling lacks custom exception classes
- Clean IPC channel design enables process isolation and component decoupling
- Absence of linting, static analysis, or testing infrastructure introduces risk of undetected bugs and style inconsistencies
What Makes It Unique
- Native integration of self-hosted LLMs like Ollama enables fully offline AI workflows without cloud dependency
- Cross-platform keychain encryption with automatic legacy credential migration ensures secure, dependency-free credential management
- Unified event-driven architecture synchronizes state across Electron, Express, and frontend frameworks via a single IPC layer
- Hybrid offline-first model with encrypted local SQLite storage and Firebase sync for seamless cloud collaboration
- Coexistence of LitElement, React, and Vue components enables progressive UI modernization while preserving performance-critical native components