@oshara/voice-sdk 0.1.0

package/dist/ui/index.d.ts

@@ -0,0 +1,21 @@
+import { VoiceAgentClient } from '../core';
+import { UIRefs } from './ui';
+export interface MountVoiceUIOptions {
+    /** Where to mount. Default document.body. */
+    target?: HTMLElement;
+    /** Fill the parent container instead of floating (hides FAB). */
+    inline?: boolean;
+    /** Open the panel on mount. */
+    openChat?: boolean;
+    /** Hide the panel close button. */
+    closeButtonHide?: boolean;
+    /** Host element id. Default "voice-agent-widget-root". */
+    rootId?: string;
+}
+export interface VoiceUIHandle {
+    destroy: () => void;
+    /** The shadow-DOM refs, for advanced host integrations. */
+    refs: UIRefs;
+}
+export declare function mountVoiceUI(client: VoiceAgentClient, opts?: MountVoiceUIOptions): VoiceUIHandle;
+export type { UIRefs } from './ui';

package/dist/ui/ui.d.ts

@@ -0,0 +1,165 @@
+import { AppearanceConfig } from '../core/appearance';
+import { FormDefinition } from '../core/forms';
+export interface FormInputBinding {
+    /** Read the current field value as a string (comma-joined for multi-select groups). */
+    read: () => string;
+    /** Toggle the disabled state across one or more underlying inputs. */
+    setDisabled: (busy: boolean) => void;
+    /** Show (or, with an empty string, clear) an inline validation error under the field. */
+    setError: (message: string) => void;
+}
+export type Screen = "welcome" | "call" | "form";
+export interface UIRefs {
+    host: HTMLDivElement;
+    shadow: ShadowRoot;
+    fab: HTMLButtonElement;
+    panel: HTMLDivElement;
+    closeBtn: HTMLButtonElement;
+    welcomeScreen: HTMLDivElement;
+    welcomeLogo: HTMLImageElement;
+    welcomeLogoFallback: HTMLDivElement;
+    welcomeName: HTMLDivElement;
+    welcomeDesc: HTMLDivElement;
+    langPicker: HTMLDivElement;
+    langTrigger: HTMLButtonElement;
+    langTriggerLabel: HTMLSpanElement;
+    langTriggerCode: HTMLSpanElement;
+    langMenu: HTMLDivElement;
+    startBtn: HTMLButtonElement;
+    startBtnLabel: HTMLSpanElement;
+    consent: HTMLDivElement;
+    consentText: HTMLSpanElement;
+    consentLink: HTMLAnchorElement;
+    poweredBy: HTMLAnchorElement;
+    /** Currently selected language code (BCP-47 short, e.g. "en"). */
+    selectedLanguage: string;
+    callScreen: HTMLDivElement;
+    callLogo: HTMLImageElement;
+    callLogoFallback: HTMLDivElement;
+    callName: HTMLDivElement;
+    callStatus: HTMLDivElement;
+    callTimer: HTMLDivElement;
+    callTimerText: HTMLSpanElement;
+    orb: HTMLDivElement;
+    orbLabel: HTMLDivElement;
+    /** Contextual processing line under the orb ("Searching the knowledge base…"). */
+    agentStatusLine: HTMLDivElement;
+    transcript: HTMLDivElement;
+    transcriptSegments: Map<string, HTMLDivElement>;
+    /**
+     * Per-role pointer to the bubble that is still in `interim` state. Used to
+     * coalesce streaming + final segments when the STT backend emits different
+     * `seg.id`s for interim vs final (e.g. sherpa), which would otherwise leave
+     * the half-finished interim bubble onscreen alongside a new final bubble.
+     */
+    transcriptInterimBubble: Map<"user" | "agent", HTMLDivElement>;
+    textInputRow: HTMLDivElement;
+    textInput: HTMLTextAreaElement;
+    textSendBtn: HTMLButtonElement;
+    muteBtn: HTMLButtonElement;
+    endBtn: HTMLButtonElement;
+    settingsBtn: HTMLButtonElement;
+    audioDrawer: HTMLDivElement;
+    audioDrawerClose: HTMLButtonElement;
+    audioMeterBar: HTMLDivElement;
+    audioMicSelect: HTMLSelectElement;
+    audioSpeakerSelect: HTMLSelectElement;
+    audioSpeakerRow: HTMLLabelElement;
+    audioVolume: HTMLInputElement;
+    audioVolumeValue: HTMLSpanElement;
+    audioNcEngine: HTMLSelectElement;
+    audioDfStrengthRow: HTMLLabelElement;
+    audioDfStrength: HTMLInputElement;
+    audioDfStrengthValue: HTMLSpanElement;
+    audioTogAec: HTMLInputElement;
+    audioTogNs: HTMLInputElement;
+    audioTogAgc: HTMLInputElement;
+    audioTogVi: HTMLInputElement;
+    audioTogViRow: HTMLLabelElement;
+    audioTogHp: HTMLInputElement;
+    audioTogTranscription: HTMLInputElement;
+    audioTogTextInput: HTMLInputElement;
+    audioDiagEngine: HTMLElement;
+    audioDiagAec: HTMLElement;
+    audioDiagNs: HTMLElement;
+    audioDiagAgc: HTMLElement;
+    audioDiagVi: HTMLElement;
+    audioDiagSr: HTMLElement;
+    audioDiagLoss: HTMLElement;
+    audioDiagJitter: HTMLElement;
+    audioDiagRtt: HTMLElement;
+    formScreen: HTMLDivElement;
+    formTranscript: HTMLDivElement;
+    formTranscriptSegments: Map<string, HTMLDivElement>;
+    formTranscriptInterimBubble: Map<"user" | "agent", HTMLDivElement>;
+    formTitle: HTMLDivElement;
+    formEyebrow: HTMLDivElement;
+    formSubtitle: HTMLDivElement;
+    formStepper: HTMLDivElement;
+    formFields: HTMLDivElement;
+    formError: HTMLDivElement;
+    formSuccess: HTMLDivElement;
+    formSubmitBtn: HTMLButtonElement;
+    formBackBtn: HTMLButtonElement;
+    formCancelBtn: HTMLButtonElement;
+    formStepBackBtn: HTMLButtonElement;
+    formCallControls: HTMLDivElement;
+    formMuteBtn: HTMLButtonElement;
+    formEndBtn: HTMLButtonElement;
+    formInputs: Map<string, FormInputBinding>;
+    appearance: AppearanceConfig;
+}
+export declare function buildUI(rootId: string, opts?: {
+    inline?: boolean;
+    parent?: HTMLElement;
+    closeButtonHide?: boolean;
+}): UIRefs;
+export declare function applyAppearance(refs: UIRefs, appearance: AppearanceConfig): void;
+export declare function setScreen(refs: UIRefs, screen: Screen): void;
+export interface RenderFormArgs {
+    definition: FormDefinition;
+    values: Record<string, string>;
+    /** Label for the cancel/back button — "Back to call" while connected, etc. */
+    cancelLabel?: string;
+    /** Which step (0-indexed) to render; ignored when the form has no `steps`. */
+    stepIndex?: number;
+    /** Fired whenever the user edits a field. Used to push state to the agent. */
+    onFieldChange?: () => void;
+}
+/**
+ * Render a form definition's fields into the form screen. Existing values are
+ * read from `values`; the on-screen inputs are kept in `refs.formInputs` so
+ * the controller can pull current values back out at submit time.
+ */
+export declare function renderForm(refs: UIRefs, args: RenderFormArgs): void;
+/** Apply inline validation errors, then clear any field not in the list. */
+export declare function setFieldErrors(refs: UIRefs, errors: {
+    name: string;
+    message: string;
+}[]): void;
+/** Clear every inline field error. */
+export declare function clearFieldErrors(refs: UIRefs): void;
+export declare function readFormValues(refs: UIRefs): Record<string, string>;
+export declare function setFormBusy(refs: UIRefs, busy: boolean): void;
+export declare function setFormError(refs: UIRefs, message: string): void;
+export declare function setFormSuccess(refs: UIRefs, message: string): void;
+export declare function setCallStatus(refs: UIRefs, text: string): void;
+export declare function setCallTimer(refs: UIRefs, remainingMs: number | null): void;
+export type OrbState = "idle" | "listening" | "speaking" | "connecting" | "thinking";
+export declare function setOrbState(refs: UIRefs, state: OrbState): void;
+/**
+ * Write the contextual processing text (e.g. "Searching the knowledge base…")
+ * under the orb, or clear it when `text` is null/empty. This is the per-tool
+ * label from a voice.agent_status event; the generic orb label still reads
+ * "Thinking…".
+ */
+export declare function setAgentStatusLine(refs: UIRefs, text: string | null): void;
+export declare function setMuteVisual(refs: UIRefs, muted: boolean): void;
+export declare function setFormCallControlsVisible(refs: UIRefs, visible: boolean): void;
+export declare function setAudioDrawerOpen(refs: UIRefs, open: boolean): void;
+export declare function setAudioMeterLevel(refs: UIRefs, level: number): void;
+export declare function populateAudioDeviceSelect(select: HTMLSelectElement, devices: MediaDeviceInfo[], selectedDeviceId: string, defaultLabel: string): void;
+export declare function appendOrUpdateTranscriptSegment(refs: UIRefs, role: "user" | "agent", segmentId: string, text: string, isFinal: boolean): void;
+export declare function clearTranscript(refs: UIRefs): void;
+export declare function showAgentTransition(refs: UIRefs, agentName: string): void;
+export declare function setFormTranscriptVisible(refs: UIRefs, visible: boolean): void;