Audio captureBROWSER-DEPENDENT
Speech recognitionBROWSER-DEPENDENT (Web Speech API, Chrome/Edge only)
Wake-word detectionSIMULATED (text prefix match, not acoustic detection)
Model routingCLOUD (InvokeLLM gateway, vendor-agnostic)
Text transformationCLOUD (rewrite) + SIMPLIFIED-LOCAL (understand)
Command executionLOCAL (explicit allowlist only)
Browser integrationBROWSER-DEPENDENT (SpeechSynthesis API)
Editor integrationSIMULATED (in-app textarea, not a real editor connection)
History storageCLOUD-HOSTED (Base44 DB, append-only, signed)
Accessibility sound alertsLOCAL analysis (Web Audio API) + BROWSER-DEPENDENT capture
None of these is yet the on-device Local GENIE Core — see GENIE_ARCHITECTURE.md.