Real-time AI-powered Bible verse detection for live sermons and broadcasts. A Tauri v2 desktop app with a React frontend and Rust backend.

Rhema listens to a live sermon audio feed, transcribes speech in real time, detects Bible verse references (both explicit citations and quoted passages), and renders them as broadcast-ready overlays via NDI for live production.

• Real-time speech-to-text via local Whisper or cloud Deepgram (WebSocket streaming + REST fallback)
  • Whisper setup guide — Local, offline transcription (no API costs)
• Multi-strategy verse detection
  • Direct reference parsing (Aho-Corasick automaton + fuzzy matching)
  • Semantic search (Qwen3-0.6B ONNX embeddings + HNSW vector index)
  • Quotation matching against known verse text
  • Cloud booster (optional, OpenAI/Claude)
  • Sermon context tracking and sentence buffering
• SQLite Bible database with FTS5 full-text search
• Multiple translations — KJV, NIV, ESV, NASB, NKJV, NLT, AMP + Spanish, French, Portuguese
• Cross-reference lookup (340k+ refs from openbible.info)
• NDI broadcast output for live production integration
• Theme designer — visual canvas editor for verse overlays with backgrounds (solid, gradient, image), text styling, positioning, shadows, and outlines
• Verse queue with drag-and-drop ordering
• Fuzzy contextual search (Fuse.js client-side)
• Audio level metering, live indicator, and session timer
• Remote control via OSC and HTTP API for hardware controllers and automation
  • Remote control guide — Stream Deck, TouchOSC, REST API integration

• Bun (runtime for scripts + package manager)
• Rust toolchain (stable, 1.77.2+)
• Tauri v2 prerequisites (platform-specific system dependencies)
• Python 3 (for downloading copyrighted translations and embedding model export)
• CMake (for Whisper local transcription) — install via brew install cmake
• Deepgram API key (optional, for cloud speech-to-text instead of Whisper)

git clone <repo-url>
cd rhema
bun install

One command sets up everything — Python virtual environment, Bible data, copyrighted translations, database, ONNX model, and precomputed embeddings:

bun run setup:all

This runs 7 phases in sequence, skipping any that are already complete:

1. Python environment setup (.venv + all pip dependencies)
2. Download open-source Bible data (KJV, Spanish, French, Portuguese + cross-references)
3. Download copyrighted translations from BibleGateway (NIV, ESV, NASB, NKJV, NLT, AMP)
4. Build SQLite Bible database (data/rhema.db with FTS5 + cross-references)
5. Download & export ONNX model (Qwen3-Embedding-0.6B) + INT8 quantization
6. Export KJV verses to JSON for embedding
7. Precompute verse embeddings (auto-selects GPU if available, falls back to ONNX CPU)

Rhema supports two speech-to-text engines:

Option 1: Whisper (Local, Free) No setup required! Whisper runs locally on your machine with no API costs or internet dependency.

• Requires CMake: brew install cmake
• Model downloads automatically on first use

Option 2: Deepgram (Cloud, Paid) Create a .env file in the project root:

DEEPGRAM_API_KEY=your_key_here

Get your API key at deepgram.com

For broadcast output via NDI:

bun run download:ndi-sdk

Each phase can also be run independently:

bun run download:bible-data          # Public domain translations + cross-refs
python3 data/download-biblegateway.py  # Copyrighted translations (needs .venv)
bun run build:bible                  # Build SQLite database
bun run download:model               # Download & export ONNX model
bun run export:verses                # Export verses to JSON
python3 data/precompute-embeddings.py  # Precompute embeddings (GPU or ONNX fallback)

bun run tauri dev

bun run tauri build

rhema/
├── src/                          # React frontend
│   ├── components/
│   │   ├── broadcast/            # Theme designer, NDI settings
│   │   ├── controls/             # Transport bar
│   │   ├── layout/               # Dashboard layout
│   │   ├── panels/               # Transcript, preview, live output, queue, search, detections
│   │   └── ui/                   # shadcn/ui + custom components
│   ├── hooks/                    # useAudio, useTranscription, useDetection, useBible, useBroadcast
│   ├── stores/                   # Zustand stores (audio, transcript, bible, queue, detection, broadcast, settings)
│   ├── types/                    # TypeScript type definitions
│   └── lib/                      # Context search (Fuse.js), verse renderer (Canvas 2D), builtin themes
├── src-tauri/                    # Rust backend (Tauri v2)
│   ├── crates/
│   │   ├── audio/                # Audio capture & metering (cpal)
│   │   ├── stt/                  # Deepgram STT (WebSocket + REST)
│   │   ├── bible/                # SQLite Bible DB, search, cross-references
│   │   ├── detection/            # Verse detection pipeline
│   │   │   ├── direct/           # Aho-Corasick + fuzzy reference parsing
│   │   │   └── semantic/         # ONNX embeddings, HNSW index, cloud booster, ensemble
│   │   ├── broadcast/            # NDI output (FFI)
│   │   ├── api/                  # Tauri command layer
│   │   └── notes/                # (placeholder)
│   └── tauri.conf.json
├── data/                         # Bible data pipeline
│   ├── prepare-embeddings.ts     # Unified setup orchestrator (bun run setup:all)
│   ├── lib/python-env.ts         # Shared Python venv management utilities
│   ├── download-sources.ts       # Download public domain translations + cross-refs
│   ├── download-biblegateway.py  # Download copyrighted translations (NIV, ESV, etc.)
│   ├── build-bible-db.ts         # Build SQLite DB from JSON sources
│   ├── compute-embeddings.ts     # Export verses to JSON for embedding
│   ├── precompute-embeddings.py  # Precompute embeddings (GPU auto-detect, ONNX fallback)
│   ├── download-model.ts         # Export & quantize Qwen3 ONNX model
│   ├── download-ndi-sdk.ts       # Download NDI SDK libraries
│   └── schema.sql                # Database schema
├── models/                       # ML models (gitignored)
├── embeddings/                   # Precomputed vectors (gitignored)
├── sdk/ndi/                      # NDI SDK files (downloaded)
└── build/                        # Vite build output

Create a .env file in the project root (optional):