Install

Pick the form that fits how you work. You can install more than one: they share nothing but the catalog, so the desktop app and a home server can live side by side.

You want…	Install
An app on your own computer	Desktop app
Tools in your terminal and scripts	CLI
One server for your household, team or agents	Docker
To hack on Dogubako itself	From source

Desktop app

The current release is 0.1.0-alpha.1, a test build. The desktop app is available for macOS 15 (Sequoia) or newer on Apple silicon. Installers for Windows, Linux and Intel Macs aren’t part of the alpha; you can build those yourself.

1. Download Dogubako_0.1.0-alpha.1_aarch64.dmg from the GitHub release, open it and drag Dogubako into Applications.
2. Open Dogubako. If the build isn’t signed yet, macOS refuses the first time; Try the alpha shows the two clicks that let it open.
3. That’s it: the common engines (FFmpeg, qpdf, Poppler, ExifTool and 7-Zip) are inside the app, and the bigger ones download with one click the first time a tool needs them. See Engines below.

CLI

The dogubako binary contains the whole engine, the web server and the MCP server. Install it from a checkout with Cargo (Rust 1.95 or newer):

git clone https://github.com/mischa-dev/dogubako && cd dogubako
cd web && npm ci && npm run build && cd ..   # optional: embeds the web UI for `server start`
cargo install --path crates/cli

Each release also has standalone CLI binaries for Linux (x86-64), macOS (Apple silicon) and Windows (x86-64), plus SHA256SUMS.txt. Put the binary on your PATH and check it:

dogubako --version
dogubako tools list

Continue with the CLI overview.

Docker

The homelab image bakes in every engine at pinned versions, so every tool works out of the box. Build it from a checkout:

git clone https://github.com/mischa-dev/dogubako && cd dogubako
cp .env.example .env            # optional: extra engines, a shared plain-words model
docker compose up -d --build
docker compose logs dogubako | grep "Setup token"

Open http://localhost:8420, paste the setup token and create the owner account. Each release also publishes the image as ghcr.io/mischa-dev/dogubako:<version>; for the alpha the package is still private, so build it yourself for now. The full guide, including Tailscale, speech recognition and a local model, is in Docker.

From source

You need Rust 1.95 or newer, a C compiler, and Node.js 20 or newer with npm.

git clone https://github.com/mischa-dev/dogubako && cd dogubako
cd web && npm ci && npm run build && cd ..
cargo build --release -p dogubako-cli
./target/release/dogubako server start      # http://127.0.0.1:8420

Building covers the desktop app, feature flags for smaller binaries and running the tests.

Engines

Dogubako calls well-known open-source programs for the heavy work. How you get them depends on what you installed:

You run	Engines
The macOS app	Managed by the app: five are built in, the rest download into the app’s folder on first use. Nothing to install
The Docker image	All baked into the image
The CLI, or a native server	Whatever is installed on the system

A tool whose engine is missing stays visible, is marked unavailable and tells you what to do. For the CLI and native servers:

dogubako deps          # what is installed, what is missing, and how to install it

Common sets:

# macOS (Homebrew)
brew install ffmpeg qpdf poppler tesseract exiftool imagemagick ghostscript sevenzip

# Debian / Ubuntu
sudo apt install ffmpeg qpdf poppler-utils tesseract-ocr libimage-exiftool-perl \
  imagemagick ghostscript 7zip

# Windows (winget)
winget install Gyan.FFmpeg QPDF.QPDF

Speech recognition, face detection and text to speech also need small local models, which you download from Settings → Models or with dogubako models download <id>. See Engines and models for the full list.

Updating

• Desktop: install the newer release over the old one. Downloaded models, extensions and outputs live in your user data folder and are kept. When an update pins a newer engine version, the app downloads it the next time it’s needed and deletes the old one.
• CLI: replace the binary, or run cargo install --path crates/cli again from an updated checkout.
• Docker: git pull && docker compose up -d --build. Your data lives in ./data and survives rebuilds. See Operations.