@alia.onl/sdk 7.2.9 → 10.0.0

package/src/components/ui/prompt-input/prompt-input.tsx

@@ -49,8 +49,6 @@ export type PromptInputProps = {
   onSuggestionUsed?: (suggestionId: string) => void;
   // Injectable error handler (replaces toast)
   onError?: (message: string) => void;
-  // Mic button API URL
-  apiUrl?: string;
 } & Omit<React.ComponentProps<typeof View>, "children">;
 
 export function PromptInput({
@@ -77,7 +75,6 @@ export function PromptInput({
   useSuggestions,
   onSuggestionUsed,
   onError,
-  apiUrl,
   ...props
 }: PromptInputProps) {
   const [internalValue, setInternalValue] = useState(value || "");
@@ -182,7 +179,7 @@ export function PromptInput({
           {actionsLeft ?? <PromptInputAddMenu />}
         </View>
         <View className="flex-row items-center gap-1.5">
-          <PromptInputMicButton apiUrl={apiUrl} />
+          <PromptInputMicButton />
           <PromptInputSubmitButton
             isLoading={isLoading}
             onStop={onStop}

package/src/components/voice/VoiceControls.tsx

@@ -1,46 +1,63 @@
 /**
- * Bottom voice control bar that replaces PromptInput when voice mode is active.
- * Contains status text, mute/cohost/end buttons, and "Continue" for cohost rounds.
+ * Bottom voice control bar that replaces the composer when voice mode is active.
+ * Contains status text and mute/end buttons.
  */
 
 import { View, Text, Pressable, ActivityIndicator, StyleSheet } from 'react-native';
 import Mic from 'lucide-react-native/icons/mic';
 import MicOff from 'lucide-react-native/icons/mic-off';
 import PhoneOff from 'lucide-react-native/icons/phone-off';
-import Users from 'lucide-react-native/icons/users';
 import type { RoomState, AgentState } from '../../types';
 
-interface VoiceControlsProps {
+/** Every word the bar shows or announces. An app in another language passes its own. */
+export interface VoiceControlsLabels {
+  connecting: string;
+  connected: string;
+  listening: string;
+  muted: string;
+  thinking: string;
+  speaking: string;
+  mute: string;
+  unmute: string;
+  end: string;
+}
+
+export const VOICE_CONTROLS_LABELS: VoiceControlsLabels = {
+  connecting: 'Connecting...',
+  connected: 'Connected',
+  listening: 'Listening...',
+  muted: 'Muted',
+  thinking: 'Thinking...',
+  speaking: 'Speaking...',
+  mute: 'Mute',
+  unmute: 'Unmute',
+  end: 'End',
+};
+
+export interface VoiceControlsProps {
   roomState: RoomState;
   agentState: AgentState;
   isMuted: boolean;
-  cohostActive: boolean;
-  currentSpeaker: 'primary' | 'cohost' | 'user' | null;
-  roundComplete: boolean;
   onToggleMute: () => void;
-  onEnableCohost: () => void;
-  onDisableCohost: () => void;
-  onContinueCohost: () => void;
   onEnd: () => void;
   /** Override theme primary color (resolved hex/hsl value, not CSS var) */
   primaryColor?: string;
+  /** Overrides for any of {@link VOICE_CONTROLS_LABELS}. */
+  labels?: Partial<VoiceControlsLabels>;
 }
 
 function getStatusText(
   roomState: RoomState,
   agentState: AgentState,
   isMuted: boolean,
-  cohostActive: boolean,
-  currentSpeaker: string | null,
+  labels: VoiceControlsLabels,
 ): string {
-  if (roomState === 'connecting') return 'Connecting...';
+  if (roomState === 'connecting') return labels.connecting;
   if (roomState === 'connected') {
-    if (cohostActive && currentSpeaker === 'cohost') return 'Cohost speaking...';
-    if (cohostActive && currentSpeaker === 'primary') return 'Alia speaking...';
-    if (agentState === 'listening') return isMuted ? 'Muted' : 'Listening...';
-    if (agentState === 'thinking') return 'Thinking...';
-    if (agentState === 'speaking') return 'Speaking...';
-    return 'Connected';
+    if (agentState === 'listening') return isMuted ? labels.muted : labels.listening;
+    if (agentState === 'thinking') return labels.thinking;
+    if (agentState === 'speaking') return labels.speaking;
+    return labels.connected;
   }
   return '';
 }
@@ -49,43 +66,31 @@ export function VoiceControls({
   roomState,
   agentState,
   isMuted,
-  cohostActive,
-  currentSpeaker,
-  roundComplete,
   onToggleMute,
-  onEnableCohost,
-  onDisableCohost,
-  onContinueCohost,
   onEnd,
   primaryColor,
+  labels: labelOverrides,
 }: VoiceControlsProps) {
-  const statusText = getStatusText(roomState, agentState, isMuted, cohostActive, currentSpeaker);
+  const labels = { ...VOICE_CONTROLS_LABELS, ...labelOverrides };
+  const statusText = getStatusText(roomState, agentState, isMuted, labels);
 
   return (
     <View style={styles.container}>
       {statusText ? (
-        <Text className="text-lg font-medium mb-4 text-foreground">
+        <Text className="text-lg font-medium mb-4 text-foreground" accessibilityLiveRegion="polite" aria-live="polite">
           {statusText}
         </Text>
       ) : null}
 
-      {roundComplete && (
-        <Pressable
-          onPress={onContinueCohost}
-          className="bg-primary/30"
-          style={[styles.continueButton, primaryColor ? { backgroundColor: primaryColor + '4D' } : undefined]}
-        >
-          <Text className="text-sm font-medium text-primary" style={primaryColor ? { color: primaryColor } : undefined}>
-            Continue conversation
-          </Text>
-        </Pressable>
-      )}
-
       {roomState === 'connected' && (
         <View style={styles.buttonRow}>
           <View style={styles.buttonWrapper}>
             <Pressable
               onPress={onToggleMute}
+              accessibilityRole="button"
+              accessibilityLabel={labels.mute}
+              accessibilityState={{ selected: isMuted }}
+              aria-pressed={isMuted}
               className={isMuted ? undefined : 'bg-muted'}
               style={[styles.button, isMuted ? { backgroundColor: '#ef4444' } : undefined]}
             >
@@ -96,34 +101,20 @@ export function VoiceControls({
               )}
             </Pressable>
             <Text className="text-xs text-muted-foreground">
-              {isMuted ? 'Unmute' : 'Mute'}
-            </Text>
-          </View>
-
-          <View style={styles.buttonWrapper}>
-            <Pressable
-              onPress={cohostActive ? onDisableCohost : onEnableCohost}
-              className={cohostActive ? 'bg-primary' : 'bg-muted'}
-              style={[
-                styles.button,
-                cohostActive && primaryColor ? { backgroundColor: primaryColor } : undefined,
-              ]}
-            >
-              <Users size={24} color="white" />
-            </Pressable>
-            <Text className="text-xs text-muted-foreground">
-              {cohostActive ? 'Solo' : 'Cohost'}
+              {isMuted ? labels.unmute : labels.mute}
             </Text>
           </View>
 
           <View style={styles.buttonWrapper}>
             <Pressable
               onPress={onEnd}
+              accessibilityRole="button"
+              accessibilityLabel={labels.end}
               style={[styles.button, { backgroundColor: '#ef4444' }]}
             >
               <PhoneOff size={24} color="white" />
             </Pressable>
-            <Text className="text-xs text-muted-foreground">End</Text>
+            <Text className="text-xs text-muted-foreground">{labels.end}</Text>
           </View>
         </View>
       )}
@@ -157,10 +148,4 @@ const styles = StyleSheet.create({
     alignItems: 'center',
     justifyContent: 'center',
   },
-  continueButton: {
-    marginBottom: 16,
-    paddingHorizontal: 20,
-    paddingVertical: 8,
-    borderRadius: 24,
-  },
 });

package/src/components/voice/VoiceSession.tsx

@@ -1,13 +1,12 @@
 /**
- * A live LiveKit call, mounted only while voice mode is on.
+ * A live voice call, mounted only while voice mode is on.
  *
- * This is the only place the chat UI touches `livekit-client`. The chat shell
- * receives it as the `voiceSession` prop rather than importing it, so a
- * text-only consumer never pulls the LiveKit client into its module graph —
- * which is the whole reason the `@alia.onl/sdk/voice` entry exists.
+ * The chat shell receives it as the `voiceSession` prop rather than importing
+ * it, so a text-only consumer never compiles the voice loop, speech playback or
+ * the call UI — which is the reason the `@alia.onl/sdk/voice` entry exists.
  *
- * Mounting starts the call; unmounting ends it (`useVoiceRoom` tears the room
- * down on unmount), so the shell only has to decide whether to render it.
+ * Mounting starts the call; unmounting ends it (`useVoiceRoom` stops listening
+ * and speaking on unmount), so the shell only has to decide whether to render it.
  */
 
 import { useEffect, useRef } from 'react';
@@ -25,15 +24,9 @@ export function VoiceSession({ apiUrl, onStateChange, onEnd }: VoiceSessionProps
     isMuted,
     error,
     messages,
-    cohostActive,
-    currentSpeaker,
-    roundComplete,
     isConnected,
     connect,
     toggleMute,
-    enableCohost,
-    disableCohost,
-    continueCohost,
   } = useVoiceRoom({ apiUrl });
 
   const { captureLevel, playbackLevel } = useAudioLevelMonitor(room, isConnected);
@@ -75,13 +68,7 @@ export function VoiceSession({ apiUrl, onStateChange, onEnd }: VoiceSessionProps
       roomState={roomState}
       agentState={agentState}
       isMuted={isMuted}
-      cohostActive={cohostActive}
-      currentSpeaker={currentSpeaker}
-      roundComplete={roundComplete}
       onToggleMute={toggleMute}
-      onEnableCohost={enableCohost}
-      onDisableCohost={disableCohost}
-      onContinueCohost={continueCohost}
       onEnd={onEnd}
     />
   );

package/src/hooks/useAliaChat.ts

@@ -3,22 +3,36 @@ import { useOxy } from '@oxy.so/services';
 import type { ChatMessage, ToolInvocation } from '../types';
 import { getTextFromContent } from '../lib/content-utils';
 import { resolveModelId } from '../lib/catalogue';
-import { PREFERRED_CHAT_MODEL_ID } from '../lib/config';
 import { streamAliaChat } from '../lib/chat-transport';
 import type { AliaChatStreamEvent } from '../lib/chat-stream';
 
 const API_URL = process.env.EXPO_PUBLIC_ALIA_API_URL ?? 'https://api.alia.onl';
 
 export interface UseAliaChatOptions {
-  /** Alia API base URL (default: EXPO_PUBLIC_ALIA_API_URL or https://api.alia.onl) */
+  /**
+   * Base URL the hook sends its requests to. Defaults to
+   * `EXPO_PUBLIC_ALIA_API_URL`, then `https://api.alia.onl`.
+   *
+   * The hook appends its own paths — `POST /v1/chat/completions` for a turn and
+   * `GET /catalogue` for the model check — so the value is a host, not a route.
+   *
+   * Point this at YOUR OWN backend to use the SDK from an origin Alia does not
+   * enumerate: Alia's product route, `POST /alia/chat`, answers CORS preflights
+   * only for Alia's own origins, while a server-to-server call carries no
+   * `Origin` header and is answered on its credential alone. Your backend
+   * receives the SDK's request, forwards the `Authorization` header it carries
+   * (the signed-in user's Oxy access token) to `/alia/chat`, and streams the
+   * SSE body back unchanged. The README's *consumer-backend path* section has
+   * the minimal relay; issue #244 records why this is the supported shape today.
+   */
   apiUrl?: string;
   /**
-   * Kaana routing profile to use.
+   * The model to answer with, as `publisher/model` from `GET /catalogue`.
    *
-   * Optional, and checked against `GET /catalogue` before a request carries it:
-   * an identifier the server no longer offers is replaced rather than sent, so a
-   * retirement does not turn into a 400 inside a consumer's app. Omitted, the
-   * build's `PREFERRED_CHAT_MODEL_ID` is what gets checked.
+   * Optional, and checked against the catalogue before a request carries it: an
+   * identifier the server no longer offers is not sent, so a retirement does not
+   * turn into a 400 inside a consumer's app. Omitted, the request carries no
+   * `model` and the server's default model answers.
    */
   model?: string;
   /** App context injected as system message so Alia knows which app the user is in. */
@@ -100,7 +114,7 @@ function waitWithAbort<T>(promise: Promise<T>, signal: AbortSignal): Promise<T>
 export function useAliaChat(options: UseAliaChatOptions = {}): UseAliaChatReturn {
   const {
     apiUrl = API_URL,
-    model = PREFERRED_CHAT_MODEL_ID,
+    model,
     clientContext,
     accessToken: accessTokenProp,
   } = options;
@@ -221,7 +235,7 @@ export function useAliaChat(options: UseAliaChatOptions = {}): UseAliaChatReturn
       const trimmed = text.trim();
       if (!trimmed) return;
 
-      const activeToken = oxyServices.getAccessToken();
+      const activeToken = oxyServices.session.accessToken;
       if (accessTokenProp !== undefined && accessTokenProp !== activeToken) {
         setError('The supplied token is not the active Oxy session.');
         return;
@@ -362,7 +376,7 @@ export function useAliaChat(options: UseAliaChatOptions = {}): UseAliaChatReturn
 
       try {
         const effectiveModel = await waitWithAbort(
-          resolveModelId(apiUrl, model, undefined, PREFERRED_CHAT_MODEL_ID),
+          resolveModelId(apiUrl, model),
           controller.signal,
         );
         await streamAliaChat(

package/src/hooks/useAudioLevelMonitor.ts

@@ -1,141 +1,45 @@
 /**
- * Real-time audio level monitoring using Web Audio API.
+ * A voice call's live levels, as React state at ~20fps.
  *
- * Extracts RMS levels from LiveKit audio tracks (local mic + remote agents)
- * and returns them at ~20fps for driving wave animations.
- *
- * Supports multiple remote audio tracks (primary + cohost) by taking
- * the max RMS level across all subscribed remote tracks.
+ * `useVoiceRoom` writes the microphone level (from the recognizer) and the
+ * playback level (from the clip being spoken) into a `VoiceLevelSource`; this
+ * reads them on a clock so a wave can follow them without rendering once per
+ * audio buffer. It used to put Web Audio analysers on a LiveKit room's tracks;
+ * the signature kept its shape — `room` is now the level source `useVoiceRoom`
+ * returns under that name.
  */
 
-import { useState, useEffect, useRef } from 'react';
-import { Room, RoomEvent, Track, type RemoteTrack, type RemoteTrackPublication, type RemoteParticipant } from 'livekit-client';
+import { useState, useEffect } from 'react';
+import type { VoiceLevelSource } from '../lib/voice-levels';
 
-/** Compute RMS (root mean square) amplitude from an AnalyserNode. Returns 0-1. */
-function getRMS(analyser: AnalyserNode, buffer: Float32Array<ArrayBuffer>): number {
-  analyser.getFloatTimeDomainData(buffer);
-  let sum = 0;
-  for (let i = 0; i < buffer.length; i++) sum += buffer[i] * buffer[i];
-  return Math.min(1, Math.sqrt(sum / buffer.length) * 8);
-}
+const POLL_MS = 50;
 
-export function useAudioLevelMonitor(room: Room | null, isConnected: boolean) {
+export function useAudioLevelMonitor(room: VoiceLevelSource | null, isConnected: boolean) {
   const [captureLevel, setCaptureLevel] = useState(0);
   const [playbackLevel, setPlaybackLevel] = useState(0);
-  const cleanupRef = useRef<(() => void) | null>(null);
 
   useEffect(() => {
-    // Tear down previous monitors
-    cleanupRef.current?.();
-    cleanupRef.current = null;
-
     if (!room || !isConnected) {
       setCaptureLevel(0);
       setPlaybackLevel(0);
-      return;
-    }
-
-    // Guard: Web Audio API required
-    if (typeof AudioContext === 'undefined') return;
-
-    const audioCtx = new AudioContext();
-    // Resume AudioContext (may be suspended until user gesture on mobile/WebView)
-    if (audioCtx.state === 'suspended') {
-      audioCtx.resume().catch(() => {});
-    }
-
-    let localAnalyser: AnalyserNode | null = null;
-    let localBuffer: Float32Array<ArrayBuffer> | null = null;
-    const remoteAnalysers = new Map<string, { analyser: AnalyserNode; buffer: Float32Array<ArrayBuffer> }>();
-    let rafId = 0;
-    let disposed = false;
-
-    // ---------- Local mic ----------
-    const localTrack = room.localParticipant?.getTrackPublication(Track.Source.Microphone)?.track;
-    if (localTrack?.mediaStreamTrack) {
-      try {
-        const source = audioCtx.createMediaStreamSource(new MediaStream([localTrack.mediaStreamTrack]));
-        localAnalyser = audioCtx.createAnalyser();
-        localAnalyser.fftSize = 256;
-        source.connect(localAnalyser);
-        localBuffer = new Float32Array(localAnalyser.fftSize);
-      } catch {
-        // MediaStream creation can fail in some environments
-      }
-    }
-
-    // ---------- Remote agent audio (supports multiple tracks) ----------
-    const setupRemoteAnalyser = (mediaStreamTrack: MediaStreamTrack, key: string) => {
-      if (remoteAnalysers.has(key) || disposed) return;
-      try {
-        const source = audioCtx.createMediaStreamSource(new MediaStream([mediaStreamTrack]));
-        const analyser = audioCtx.createAnalyser();
-        analyser.fftSize = 256;
-        source.connect(analyser);
-        const buffer = new Float32Array(analyser.fftSize);
-        remoteAnalysers.set(key, { analyser, buffer });
-      } catch {
-        // Ignore errors
-      }
-    };
-
-    // Check already-subscribed remote tracks
-    for (const [, p] of room.remoteParticipants) {
-      for (const [, pub] of p.audioTrackPublications) {
-        if (pub.track?.mediaStreamTrack) {
-          setupRemoteAnalyser(pub.track.mediaStreamTrack, `${p.identity}-${pub.trackSid}`);
-        }
-      }
+      return undefined;
     }
 
-    // Listen for new remote audio tracks
-    const onTrackSubscribed = (track: RemoteTrack, publication: RemoteTrackPublication, participant: RemoteParticipant) => {
-      if (track.kind === Track.Kind.Audio && track.mediaStreamTrack) {
-        setupRemoteAnalyser(track.mediaStreamTrack, `${participant.identity}-${publication.trackSid}`);
-      }
-    };
-    room.on(RoomEvent.TrackSubscribed, onTrackSubscribed);
-
-    // Clean up stale analysers when tracks are unsubscribed
-    const onTrackUnsubscribed = (_track: RemoteTrack, publication: RemoteTrackPublication, participant: RemoteParticipant) => {
-      const key = `${participant.identity}-${publication.trackSid}`;
-      remoteAnalysers.delete(key);
+    let dirty = true;
+    const unsubscribe = room.subscribe(() => {
+      dirty = true;
+    });
+    const timer = setInterval(() => {
+      if (!dirty) return;
+      dirty = false;
+      setCaptureLevel(room.capture);
+      setPlaybackLevel(room.playback);
+    }, POLL_MS);
+
+    return () => {
+      clearInterval(timer);
+      unsubscribe();
     };
-    room.on(RoomEvent.TrackUnsubscribed, onTrackUnsubscribed);
-
-    // ---------- Poll loop (~20fps) ----------
-    let lastUpdate = 0;
-    const poll = (time: number) => {
-      if (disposed) return;
-      if (time - lastUpdate > 50) {
-        lastUpdate = time;
-        if (localAnalyser && localBuffer) {
-          setCaptureLevel(getRMS(localAnalyser, localBuffer));
-        }
-        // Take max RMS across all remote tracks (primary + cohost)
-        let maxRemoteLevel = 0;
-        for (const [, { analyser, buffer }] of remoteAnalysers) {
-          const level = getRMS(analyser, buffer);
-          if (level > maxRemoteLevel) maxRemoteLevel = level;
-        }
-        setPlaybackLevel(maxRemoteLevel);
-      }
-      rafId = requestAnimationFrame(poll);
-    };
-    rafId = requestAnimationFrame(poll);
-
-    // ---------- Cleanup ----------
-    const dispose = () => {
-      disposed = true;
-      cancelAnimationFrame(rafId);
-      room.off(RoomEvent.TrackSubscribed, onTrackSubscribed);
-      room.off(RoomEvent.TrackUnsubscribed, onTrackUnsubscribed);
-      remoteAnalysers.clear();
-      audioCtx.close().catch(() => {});
-    };
-    cleanupRef.current = dispose;
-
-    return dispose;
   }, [room, isConnected]);
 
   return { captureLevel, playbackLevel };