AZ Droid Works Droid design and robotics lab

Home droid workshop

Building Sonny from real systems.

AZ Droid Works is a small robotics lab building Sonny, a friendly home droid whose current brain lives on a durable VM while Raspberry Pi edge modules handle the physical world: voice capture, speaker playback, expressive eyes, sensors, and eventually the mobile base.

First build

Sonny

Sonny Droid

Sonny is the first AZ Droid Works build: a practical companion-droid prototype with a VM-centered brain, MQTT messaging, local speech paths, Pi voice and face nodes, and Mission Control for operator visibility.

Build signals

BrainVM host
EdgeRaspberry Pi modules
InterfaceVoice + face
OpsMission Control

Current architecture

A droid brain on the VM, edge modules on the robot.

The source of truth is the Sonny repo and build docs; this site is the public field note.

VM Brain

The always-on brain host owns orchestration, memory context, route selection, STT/TTS, MQTT command publishing, and high-level state for voice and face behavior.

Pi Edge Modules

Raspberry Pi nodes keep hardware close to the body: microphone capture, wake/capture flow, speaker playback, animated eyes, sensors, and later base safety control.

Mission Control

The operator dashboard exists so Sonny can be observed, tested, and debugged without guessing what the robot thinks it is doing. Charming, but not mysterious. Mostly.

Build log

What is real right now.

Public summary, not the operational runbook.

Head / face

Serviceable v1 head direction with expressive eye displays, camera and audio openings, Pi-based face control, and operator-adjustable face states.

Voice

Wake/capture, STT, local response routing, TTS, and inline MQTT audio playback are being hardened around a low-latency home-assistant-style voice path.

Body / base

The base is still a later subsystem. The current architecture keeps drive and safety control out at the edge instead of burying it inside the conversational brain.

Lab direction

Robots with character, built one subsystem at a time.

Readable systems. Repeatable builds. Real-world behavior.

Expressive Presence

Face, voice, attention, and timing matter. A good droid should feel understandable before it feels complicated.

Modular Engineering

Brains, sensors, displays, audio, and mobility should be testable pieces with clear contracts between them.

Operator Visibility

Mission Control-style dashboards and runbooks make the robot easier to debug, improve, and trust.