What it does SUSI.AI Server is the Java backend that powers an open-source personal assistant. It handles chat and voice interactions, routes requests to external APIs for weather, music, to-do lists, and podcasts, and manages the “intelligence” and “personality” that client apps (Android, web, Linux) consume. You host it yourself, you control it.

The interesting bit The project treats skills as git-managed data: the public skill repository lives in a separate repo you clone alongside the server, and you can run a private skill repo via a bare git host on your own machine. It’s a pleasantly old-school approach to versioning conversational logic.

Key highlights

• Builds with Gradle or Ant; runs on Java 8+ (Oracle or OpenJDK)
• Armv6 support for Raspberry Pi Zero/1A/1B, though Java 8 is required there
• Self-upgrading via bin/upgrade.sh
• Docker deployment guides for AWS, Azure, GCloud, DigitalOcean, Bluemix, plus Heroku and Scalingo buttons
• Configuration layered: base conf/config.properties, overrides in data/settings/customized_config.properties
• Skills edited at https://skills.susi.ai or authored locally in a custom skill-data repo

Caveats

• No pre-built releases; you compile from source (the README claims “one minute,” your mileage may vary)
• Production deployment on api.susi.ai was “planned” and “soon” with master branch, but the README states dev branch deploys hourly — the actual production status is unclear
• Windows build instructions still reference ant jar while the main flow uses Gradle; the dual build system looks unmaintained

Verdict Worth a look if you want a hackable, self-hosted assistant and don’t mind Java boilerplate. Skip it if you need a polished, batteries-included product — this is a build-it-yourself kit with community support via Gitter and Google Groups.