Getting Started
Prerequisites
System requirements and toolchain you'll need before installing Rhema.
Rhema is a Tauri v2 desktop app, so you'll need the standard Tauri toolchain plus a Python interpreter for the data-prep pipeline. The complete list:
| Tool | Why you need it |
|---|---|
| Bun | JavaScript runtime + package manager for web/, build scripts, and the orchestrator |
| Rust (stable, 1.77.2+) | Compiles the Tauri backend and the seven workspace crates |
| Tauri v2 prerequisites | Platform-specific system libraries (WebView2 on Windows, WebKit on Linux) |
| Python 3 | Downloads copyrighted translations and exports the embedding model to ONNX |
| CMake + LLVM/libclang | Required for local Whisper STT — whisper.cpp is built from source via bindgen |
| Deepgram API key (optional) | Cloud speech-to-text, only if you don't want to run Whisper locally |
Whisper vs Deepgram
You only need one speech-to-text engine. Whisper runs locally with no API
costs but compiles whisper.cpp from source on first build. Deepgram streams
via WebSocket and bypasses the local build but needs a paid API key. See
Speech-to-text for tradeoffs.
Hardware
Rhema runs comfortably on modern laptops:
- CPU: anything from the last five years; a recent Apple Silicon or Ryzen/Intel laptop will keep the audio capture, STT, and detection pipeline well under a single core.
- RAM: 8 GB minimum, 16 GB recommended if you precompute embeddings on CPU (the GPU path uses far less memory).
- GPU: optional. The setup script auto-detects CUDA/MPS for embedding precomputation and falls back to ONNX CPU otherwise.
- Disk: ~3 GB after
setup:allfinishes (Bible data, embeddings, ONNX models).
Operating system
| OS | Status |
|---|---|
| macOS 13+ (Apple Silicon and Intel) | First-class support |
| Windows 10/11 (x86_64) | First-class support — see Platform setup for the LLVM bootstrap |
| Linux (Debian/Ubuntu/Arch) | First-class support — needs the WebKitGTK and libclang packages |
Once these are installed, head to the installation guide.