Friday

Privacy-first AI meeting assistant for local capture, live transcription, and meeting summaries. Friday is a native desktop app that records and summarizes meetings without shipping your raw audio or transcripts to the cloud. The product is built on a privacy-first local-first architecture with a Tauri desktop shell, Rust audio/transcription services, and a Next.js interface.

---

Table of Contents

• What Friday is for
• Core capabilities
• Project structure
• Architecture overview
• Requirements
• Running Friday in development
• Building from source
• Repository links
• Contributing
• License

What Friday is for

Friday is meant for teams and professionals who need accurate meeting notes while retaining full control of their data. It is designed for:

• Local-first transcription and storage
• Real-time transcript updates
• AI summaries with configurable providers
• Meeting export and editing workflows
• Stronger privacy controls than cloud-only assistants

Core capabilities

• Local-first by default
  Meetings are recorded and processed in local pipelines on your machine.
• Professional audio mixing
  The audio layer synchronizes microphone and system audio with voice-activity-based processing for transcription quality.
• Live transcripts + summaries
  See transcriptions as they are produced and generate meeting summaries after meetings end.
• Cross-platform desktop app
  Built with Tauri for macOS, Windows, and Linux targets.
• Flexible AI provider support Gemini (primary), with support for local and external providers for summary generation.
• Privacy-first defaults
  No raw audio/notes upload required for core operation.

Project structure

• frontend/
  Desktop shell, UI, and all local Rust audio/transcription glue.
• backend/
  FastAPI service for meeting persistence, summarization endpoints, and deployment helpers.
• docs/
  Operational and build documentation.

Architecture overview

Frontend (Next.js UI)  <->  Tauri command bridge  <->  Rust services
                                                           |
                                                           +-> Audio pipeline + recording
                                                           +-> Whisper integration
                                                           +-> Local file outputs

Frontend HTTP/WebSocket   <->  FastAPI backend  <->  SQLite + LLM providers
                                                           |
                                                           +-> Whisper server for transcription orchestration

Requirements

• Rust (stable toolchain)
• Node.js 20+ and pnpm
• Python 3.8+ (for backend workflows)
• On macOS: microphone + screen recording permission for full system-audio capture
• GPU support depends on platform:
  • macOS: Metal/CoreML
  • Windows/Linux: CUDA/Vulkan where available

Running Friday in development

1. Install frontend dependencies:

cd frontend
pnpm install

2. Start frontend in Tauri mode:

pnpm run tauri:dev

3. Start backend (recommended for full stack features):

cd backend
./clean_start_backend.sh   # macOS/Linux

Windows users can use the PowerShell workflow documented in backend/README.md.

Default local ports:

• Frontend dev server: 3118
• Backend API: 5167
• Whisper server: 8178

4. Backend docs are available at http://localhost:5167/docs.

If you only need to test frontend rendering, you can run pnpm run dev from frontend for the Next.js app alone.

Building from source

• Frontend scripts (from frontend):

  • pnpm run tauri:dev full app development mode
  • pnpm run tauri:dev:cpu CPU-only development build
  • pnpm run tauri:dev:metal Apple Silicon/Metal path
  • pnpm run tauri:dev:cuda CUDA path
  • pnpm run tauri:dev:vulkan Vulkan path
  • pnpm run tauri:build production app build

• Backend scripts and deployment options:

  • Docker-based and native setup instructions in backend/README.md
  • Whisper model options: tiny, base, small, medium, large-v3 and more

Repository links

• Website documentation: tinyurl.com/2nvsk8ek
• Architecture details: docs/architecture.md
• Build details: docs/BUILDING.md
• Linux build notes: docs/building_in_linux.md
• Backend setup: backend/README.md
• Frontend notes: frontend/README.md

Contributing

Contributions are welcome.

• Keep PRs focused and scoped.
• Update docs when behavior or command flow changes.
• Use existing Rust and TypeScript patterns in the codebase.
• See CONTRIBUTING.md for full project process.

License

MIT License. See LICENSE.