End-to-end pipeline on Heimdall: Whisper STT + Mycroft Precise wake word + Piper TTS. Server-side wake detection, Home Assistant integration optional.
Wake word moves to edge hardware (K210 or ESP32-S3). Audio only streams post-wake. KMODEL conversion for K210 or ESP-SR for ESP32-S3.
Documented reference hardware build. Buying guide, assembly instructions, enclosure design. Accessible to non-technical builders.
Platform
0%
Multiple hardware targets supported. Custom PCB design. Minerva as a hardware node for linnet (tone annotation in any room).
Minerva server replaced by cf_voice modules from circuitforge-core. STT/TTS via cf_voice.io, classifier via cf_voice.context.