API: ai/nlu-ui
@mosaicoo/svg-engine/ai/nlu-ui is the UI for the natural-language layer: a
ready-made command box (text + voice) bound to the NLU engine, plus the browser
Web Speech voice provider and an engine orchestrator. Depends on Angular Material.
import { SvgeNluInput } from '@mosaicoo/svg-engine/ai/nlu-ui';The command box
Section titled “The command box”| Component | Selector | What it is |
|---|---|---|
SvgeNluInput | <svge-nlu-input> | A standalone Material command box: text field + mic button + a live preview of the top candidate (label, confidence, bar), a status line, a confirm button for destructive/low-confidence matches, and a clickable alternatives list. |
Key bindings:
| Input / Output | Purpose |
|---|---|
confirmGate? | Hook called before running a destructive or low-confidence intent. |
voiceLanguage? / voiceEngine? | Choose the recognition language (e.g. pt-BR) and engine (web-speech/whisper/auto). |
nluService? / confirmDialog? | Override the NLU service or the confirmation prompt. |
resultExecuted | Emits the NluExecuteResult after each run. |
Voice services
Section titled “Voice services”| API | Description |
|---|---|
VoiceRecognitionService | A VoiceProvider backed by the browser Web Speech API — privacy-first and free, no model download. isSupported/listening/lastError signals, listen(lang?, options?) → final transcript, stop(). |
VoiceEngineService | An orchestrator exposing the same VoiceProvider surface, so you can switch between Web Speech and the on-device Whisper provider transparently: setEngine('web-speech' | 'whisper' | 'auto'), engine/whisperAvailable/availableEngines signals. Persists the choice to localStorage. |
Both support languages such as pt-BR, en-US and es-ES. auto mode tries Web
Speech first and falls back to Whisper when it’s registered (see
ai/nlu-voice-wasm).