v1.4.0: streaming audio + acoustic wake word (openWakeWord 'hey jarvis'), gateway, capturer

v1.3.1: wake word -itty/-itti matching + transcript logging (diagnose no-response)

v1.3.0: Moonshine (sherpa-onnx) STT — ~0.85s/turn vs ~4.7s, faster-whisper fallback

v1.2.0: optional ElevenLabs cloud voice (onboarding key box, local Piper fallback)

v1.1.0: wake word ('hey Chitti'), off-thread STT, tiny.en model, greet-once

onboarding: no spoken passphrase — antenna hold -> name -> face

fix: short passphrase + speak-back; English STT; settings 422

fix: onboarding waits for antenna hold + stays alive; settings routes; SDK camera

fix: headless key custody (FileKeyStore fallback) so the app boots on the robot

v1.0.0 — torch-free build (ONNX VAD, face-only recognition); installs on the Reachy Mini

v1.0.0 — user-configurable LLM endpoint + settings UI; default model gemma4-e4b

ChittiOS v1.0.0 — full app: identity + voice conversation + on-robot onboarding (vendored core + bundled voice)

Polished Space landing page (privacy-forward, onboarding explainer, light/dark)

v0.3.0 — reference-pattern idle: 60Hz set_target control loop, sine-of-time breathing; greeting 'Hi, I'm ChittiOS'

ChittiOS — fresh public Space (bypass robot install cache); body_yaw idle + 'awake' greeting

initial commit