Voxa

Open-source meeting intelligence. Local by default. Connected when useful.

Voxa is a production-grade, privacy-first desktop application for recording, transcribing, and summarizing meetings entirely on your local machine.

---

Highlights

• 100% On-Device AI: High-accuracy local speech-to-text using NVIDIA Parakeet TDT (ONNX) and OpenAI Whisper (Whisper.cpp) with full Metal / GPU hardware acceleration.
• Zero Cloud Audio Upload: Audio streams and raw recordings remain strictly on your physical machine. No third-party servers, no accounts required.
• Dual-Channel Audio Capture: Record microphone input alongside system audio (Zoom, Google Meet, Microsoft Teams, Slack Huddles, WebRTC) without intrusive bots or meeting permissions.
• Local Summarization: Generate executive summaries, key decisions, and action items using local LLMs via Ollama or embedded llama-helper.
• Structured Workspace: Clean meeting details screen featuring timestamped transcripts, synchronized audio playback, editable notes, and one-click Markdown & Plain Text export.
• Future-Ready Architecture: Designed with modular integration boundaries (src/integrations/voxbento) to bridge seamlessly with VoxBento for real-time multilingual event interpretation.

---

Architecture Overview

Voxa Desktop Architecture
┌─────────────────────────────────────────────────────────────┐
│                    Voxa Frontend (Next.js)                  │
│   Home Dashboard │ Live Meeting Monitor │ Meeting Details   │
└──────────────────────────────┬──────────────────────────────┘
                               │ Tauri IPC Bridge
┌──────────────────────────────▼──────────────────────────────┐
│                    Tauri Backend (Rust)                     │
│  ┌───────────────────────┐         ┌──────────────────────┐ │
│  │  Native Audio Engine  │         │   Local DB (SQLite)  │ │
│  │ (CoreAudio / WASAPI)  │         │   sqlx Migrations    │ │
│  └───────────┬───────────┘         └──────────────────────┘ │
│              │ PCM 16kHz                                    │
│  ┌───────────▼───────────┐         ┌──────────────────────┐ │
│  │ Transcription Engine  │         │  Zero-Cloud Tracing  │ │
│  │ (Parakeet / Whisper)  │         │ (Local Logging Only) │ │
│  └───────────────────────┘         └──────────────────────┘ │
└──────────────────────────────┬──────────────────────────────┘
                               │ Optional Integration Seam
┌──────────────────────────────▼──────────────────────────────┐
│            VoxBento Integration (src/integrations)          │
│   REST Client │ Live Caption WebSocket │ Zero Audio Ingest  │
└─────────────────────────────────────────────────────────────┘

---

Getting Started

Prerequisites

• Node.js: v20.x or v22.x+
• pnpm: v9.x or v10.x+ (corepack enable or npm install -g pnpm)
• Rust Toolchain: 1.80+ (rustup default stable)
• macOS: Xcode Command Line Tools (xcode-select --install)
• Linux: System libraries (libwebkit2gtk-4.1-dev, libssl-dev, libasound2-dev, libgtk-3-dev)
• Windows: Visual Studio C++ Build Tools & WebView2

Quick Setup

1. Clone the repository:

   git clone https://github.com/ArnavBallinCode/voxa.git
   cd voxa
   

2. Install frontend dependencies:

   cd frontend
   pnpm install
   

3. Verify the Rust and frontend toolchains:

   # In frontend/ directory:
   pnpm run typecheck
   pnpm run build
   
   # In project root:
   cargo check --manifest-path frontend/src-tauri/Cargo.toml
   

4. Launch the desktop application in development mode:

   cd frontend
   pnpm run tauri:dev
   

---

Exporting & Data Ownership

Meetings and transcripts can be exported immediately:

• Markdown (.md): Executive summary, action checklist, key decisions, and timestamped dialogue.
• Plain Text (.txt): Formatted transcript suitable for pasting into email or project tickets.
• Raw Audio: Accessible directly via the "Recording Folder" button in Meeting Details.

---

Platform Support

---

License

Voxa is open-source software licensed under the MIT License. See ATTRIBUTIONS.md for third-party acknowledgments.