What it does

SUSI.AI Android is the mobile front-end for an open-source assistant stack. It chats, it takes voice input, it shows you “skills” (think Alexa skills, but community-curated), and it can configure SUSI smart speakers on your Wi-Fi. The app mixes Java and Kotlin, leans on Retrofit and Realm, and feeds queries to a FOSSASIA-hosted server that parses them through a shared skill repository.

The interesting bit

The project is almost more interesting as an organizational artifact than as an app. It enforces strict MVP with interface contracts, packages by feature rather than layer, and auto-publishes alpha builds to Play Store on every master merge. Someone really thought about the factory floor.

Key highlights

• Roughly 50/50 Java and Kotlin split; the README cheerfully notes Kotlin has “much more advantages than Java” without specifying which
• Heavy use of established Android libraries: Retrofit, Picasso, LeakCanary, Espresso, Snowboy for hotword detection
• Geolocation-aware answers for “things nearby” queries
• Smart speaker setup flow baked in (Wi-Fi provisioning, room assignment)
• Auto-generated APK branches for both F-Droid and Play Store on every merged PR

Caveats

• The roadmap lists hotword training and full-screen voice interaction as planned, not shipped
• README mentions “intelligent answers” but never quantifies accuracy, latency, or server uptime
• Some library choices show age: ButterKnife (largely superseded by ViewBinding), Realm (SQLite/ROOM has eaten its lunch)

Verdict

Worth a look if you’re building a voice assistant client and want to see how a large open-source project enforces architecture discipline. Skip it if you need a drop-in, production-ready assistant today; SUSI’s server dependency and unclear maintenance status make it a research project with nice screenshots.