Quick Start

Two parts required

Sinsajo always needs two sides: a Server (Rust, runs the AI) and a Client (Flutter, captures audio + shows text). They connect over WebSocket ws://YOUR_SERVER_IP:8765. Run one server, connect any client to it.

Prerequisites

Side	Requirement
Server — pick one path	Docker or Rust 1.70+ or precompiled x86_64 binary (Windows/Linux)
Client — pick one path	Modern browser (Web App) or Android device or Flutter 3.x

No cloud account needed. First server start downloads the model (~2–3GB), subsequent starts are ~2–3s.

0. Clone (only if you build from source)

git clone https://github.com/lutgaru/Sinsajo.git
cd Sinsajo

Skip this if you only use Docker or precompiled binaries / Web App / APK.

1. Start a Server — pick one

Docker Compose (recommended)

Best for a Linux server / homelab. Persists models and recordings.

docker-compose.yml

services:
  sinsajo-server:
    image: lutgaru/sinsajo-server:latest
    ports: ["8765:8765"]
    volumes:
      - sinsajo_models:/app/models
      - sinsajo_records:/app/records
    restart: unless-stopped
volumes:
  sinsajo_models:
  sinsajo_records:

docker compose up -d
# -> ws://localhost:8765 (or your LAN IP:8765)

See Server Guide for tags (--model Canary180M) and volumes.

Docker CLI

Same image without a compose file:

docker pull lutgaru/sinsajo-server:latest
docker run -p 8765:8765 \
  -v sinsajo_models:/app/models \
  -v sinsajo_records:/app/records \
  lutgaru/sinsajo-server:latest

Precompiled binary

No Docker / no Rust needed. Download for your OS/arch.

# From GitHub Releases -> server/v* assets:
# - sinsajo-server        (linux x86_64)
# - sinsajo-server.exe    (windows x86_64)
./sinsajo-server --model ParakeetTDT --autodownload-model --port 8765
# Windows: .\sinsajo-server.exe --model ParakeetTDT --autodownload-model

Get it at server-latest. No compilation.

CLI args:

Arg	Default	Description
--model	prompt	ParakeetTDT or Canary180M
--autodownload-model	false	Auto-download if missing
--port	8765	WebSocket port
--model-dir	./models	Model storage
--records-dir	./records	WAV/OGG output

Build from source

For any architecture that meets minimum requirements (Rust 1.70+, ~500MB RAM for Int8).

cd server
cargo run --release -- --model ParakeetTDT --autodownload-model
# -> ws://0.0.0.0:8765
# or build: cargo build --release && ./target/release/sinsajo-server --help

If --model is omitted it reads sinsajo-config.json, then prompts.

Model download

Server auto-downloads on first run (~2–3GB from HuggingFace). Override dirs with --model-dir / --records-dir. Use Canary180M for smaller/faster multilingual, ParakeetTDT for English accuracy.

2. Connect a Client — pick one

Web App (no install)

Fastest — nothing to install.

1. Ensure your server is running (ws://YOUR_SERVER_IP:8765 reachable)
2. Open Web App (or https://lutgaru.github.io/Sinsajo/app/)
3. Settings → set server IP → tap mic → speak

Requires HTTPS for microphone (GitHub Pages provides it). See Client Guide.

Android APK

For phones/tablets — share via link or QR.

• Download app-release.apk from client-latest
• Enable Install unknown apps → install → set server IP in Settings → speak

Built on every client/v* tag via GitHub Actions. Do not commit APKs to git.

Build Flutter app

For any platform meeting client requirements (Flutter 3.x): Android, iOS, Linux, macOS, Windows, Web.

cd sinsajo_client
flutter pub get
flutter run                                # on connected device/emulator
flutter build apk --release                # Android
flutter build web --release --base-href /Sinsajo/app/  # Web (merged into docs-site/dist/app/)

Default URL is lib/providers/transcription_provider.dart → kWsUrl, overridden in Settings UI (gain, source, save_audio, target_language).

3. Use it

1. Tap the microphone button in the client
2. Speak — VAD highlights when speech is detected (only speech is sent)
3. See real-time text; model_info tells you which languages are supported
4. Tap Stop or Clean to end — clean also saves WAV/OGG on the server records/
5. Copy or clear the transcription

60-second test

docker compose up -d → server on localhost:8765 → open Web App → set localhost or LAN IP → tap mic.

See also: Architecture, Configuration, Client, Server.