@absolutejs/voice 0.0.22-beta.648 → 0.0.22-beta.649

package/dist/angular/index.js

@@ -343,29 +343,34 @@ var createMicrophoneCapture = (options) => {
     if (!AudioContextCtor) {
       throw new Error("Browser microphone capture requires AudioContext support.");
     }
-    mediaStream = options.stream ?? await navigator.mediaDevices.getUserMedia({
-      audio: {
-        autoGainControl: true,
-        channelCount: options.channelCount ?? 1,
-        echoCancellation: true,
-        noiseSuppression: true
+    try {
+      mediaStream = options.stream ?? await navigator.mediaDevices.getUserMedia({
+        audio: {
+          autoGainControl: true,
+          channelCount: options.channelCount ?? 1,
+          echoCancellation: true,
+          noiseSuppression: true
+        }
+      });
+      audioContext = new AudioContextCtor;
+      if (audioContext.state === "suspended") {
+        await audioContext.resume();
       }
-    });
-    audioContext = new AudioContextCtor;
-    if (audioContext.state === "suspended") {
-      await audioContext.resume();
-    }
-    sourceNode = audioContext.createMediaStreamSource(mediaStream);
-    processorNode = audioContext.createScriptProcessor(4096, 1, 1);
-    processorNode.onaudioprocess = (event) => {
-      const channel = event.inputBuffer.getChannelData(0);
-      const downsampled = downsampleBuffer(channel, audioContext?.sampleRate ?? 48000, options.sampleRateHz ?? 16000);
-      const pcm = floatTo16BitPCM(downsampled);
-      options.onLevel?.(getPcmLevel(pcm));
-      options.onAudio(pcm);
-    };
-    sourceNode.connect(processorNode);
-    processorNode.connect(audioContext.destination);
+      sourceNode = audioContext.createMediaStreamSource(mediaStream);
+      processorNode = audioContext.createScriptProcessor(4096, 1, 1);
+      processorNode.onaudioprocess = (event) => {
+        const channel = event.inputBuffer.getChannelData(0);
+        const downsampled = downsampleBuffer(channel, audioContext?.sampleRate ?? 48000, options.sampleRateHz ?? 16000);
+        const pcm = floatTo16BitPCM(downsampled);
+        options.onLevel?.(getPcmLevel(pcm));
+        options.onAudio(pcm);
+      };
+      sourceNode.connect(processorNode);
+      processorNode.connect(audioContext.destination);
+    } catch (error) {
+      stop();
+      throw error;
+    }
   };
   const stop = () => {
     processorNode?.disconnect();
@@ -585,6 +590,7 @@ var noopUnsubscribe = () => noop;
 var NOOP_CONNECTION = {
   callControl: noop,
   close: noop,
+  disconnect: noop,
   endTurn: noop,
   send: noop,
   sendAudio: noop,
@@ -596,7 +602,7 @@ var NOOP_CONNECTION = {
   start: () => {}
 };
 var createSessionId = () => crypto.randomUUID();
-var buildWsUrl = (path, sessionId, scenarioId) => {
+var buildWsUrl = (path, sessionId, scenarioId, query) => {
   const { hostname, port, protocol } = window.location;
   const wsProtocol = protocol === "https:" ? "wss:" : "ws:";
   const portSuffix = port ? `:${port}` : "";
@@ -605,6 +611,11 @@ var buildWsUrl = (path, sessionId, scenarioId) => {
   if (scenarioId) {
     url.searchParams.set(DEFAULT_SCENARIO_QUERY_PARAM, scenarioId);
   }
+  for (const [key, value] of Object.entries(query)) {
+    if (key !== "sessionId" && key !== DEFAULT_SCENARIO_QUERY_PARAM && value) {
+      url.searchParams.set(key, value);
+    }
+  }
   return url.toString();
 };
 var isVoiceServerMessage = (value) => {
@@ -714,7 +725,7 @@ var createVoiceConnection = (path, options = {}) => {
     }, delayMs);
   };
   const connect = () => {
-    const ws = new WebSocket(buildWsUrl(path, state.sessionId, state.scenarioId));
+    const ws = new WebSocket(buildWsUrl(path, state.sessionId, state.scenarioId, options.query ?? {}));
     ws.binaryType = "arraybuffer";
     ws.onopen = () => {
       const wasReconnecting = state.reconnectAttempts > 0;
@@ -810,7 +821,19 @@ var createVoiceConnection = (path, options = {}) => {
       type: "call_control"
     });
   };
-  const close = () => {
+  const close = (reason = "client-close") => {
+    clearTimers();
+    if (state.ws) {
+      if (state.ws.readyState === WS_OPEN) {
+        state.ws.send(JSON.stringify({ reason, type: "close" }));
+      }
+      state.ws.close(WS_NORMAL_CLOSURE);
+      state.ws = null;
+    }
+    state.isConnected = false;
+    listeners.clear();
+  };
+  const disconnect = () => {
     clearTimers();
     if (state.ws) {
       state.ws.close(WS_NORMAL_CLOSURE);
@@ -834,6 +857,7 @@ var createVoiceConnection = (path, options = {}) => {
   return {
     callControl,
     close,
+    disconnect,
     endTurn,
     send,
     sendAudio,
@@ -1114,10 +1138,17 @@ var createVoiceStream = (path, options = {}) => {
     callControl(message) {
       connection.callControl(message);
     },
-    close() {
+    close(reason) {
       unsubscribeConnection();
       browserMediaReporter?.close();
-      connection.close();
+      connection.close(reason);
+      store.dispatch({ type: "disconnected" });
+      notify();
+    },
+    disconnect() {
+      unsubscribeConnection();
+      browserMediaReporter?.close();
+      connection.disconnect();
       store.dispatch({ type: "disconnected" });
       notify();
     },
@@ -1599,6 +1630,7 @@ var createVoiceController = (path, options = {}) => {
       };
       notify();
     } catch (error) {
+      capture?.stop();
       capture = null;
       state = {
         ...state,
@@ -1609,13 +1641,19 @@ var createVoiceController = (path, options = {}) => {
       throw error;
     }
   };
-  const close = () => {
+  const close = (reason) => {
+    unsubscribeStream();
+    stopRecording();
+    stream.close(reason);
+  };
+  const disconnect = () => {
     unsubscribeStream();
     stopRecording();
-    stream.close();
+    stream.disconnect();
   };
   return {
     close,
+    disconnect,
     startRecording,
     stopRecording,
     get assistantAudio() {

package/dist/client/connection.d.ts

@@ -11,7 +11,8 @@ type VoiceConnectionHandle = {
         sessionId?: string;
         scenarioId?: string;
     }) => void;
-    close: () => void;
+    close: (reason?: string) => void;
+    disconnect: () => void;
     endTurn: () => void;
     getReadyState: () => number;
     getScenarioId: () => string;

package/dist/client/htmxBootstrap.js

@@ -111,29 +111,34 @@ var createMicrophoneCapture = (options) => {
     if (!AudioContextCtor) {
       throw new Error("Browser microphone capture requires AudioContext support.");
     }
-    mediaStream = options.stream ?? await navigator.mediaDevices.getUserMedia({
-      audio: {
-        autoGainControl: true,
-        channelCount: options.channelCount ?? 1,
-        echoCancellation: true,
-        noiseSuppression: true
+    try {
+      mediaStream = options.stream ?? await navigator.mediaDevices.getUserMedia({
+        audio: {
+          autoGainControl: true,
+          channelCount: options.channelCount ?? 1,
+          echoCancellation: true,
+          noiseSuppression: true
+        }
+      });
+      audioContext = new AudioContextCtor;
+      if (audioContext.state === "suspended") {
+        await audioContext.resume();
       }
-    });
-    audioContext = new AudioContextCtor;
-    if (audioContext.state === "suspended") {
-      await audioContext.resume();
-    }
-    sourceNode = audioContext.createMediaStreamSource(mediaStream);
-    processorNode = audioContext.createScriptProcessor(4096, 1, 1);
-    processorNode.onaudioprocess = (event) => {
-      const channel = event.inputBuffer.getChannelData(0);
-      const downsampled = downsampleBuffer(channel, audioContext?.sampleRate ?? 48000, options.sampleRateHz ?? 16000);
-      const pcm = floatTo16BitPCM(downsampled);
-      options.onLevel?.(getPcmLevel(pcm));
-      options.onAudio(pcm);
-    };
-    sourceNode.connect(processorNode);
-    processorNode.connect(audioContext.destination);
+      sourceNode = audioContext.createMediaStreamSource(mediaStream);
+      processorNode = audioContext.createScriptProcessor(4096, 1, 1);
+      processorNode.onaudioprocess = (event) => {
+        const channel = event.inputBuffer.getChannelData(0);
+        const downsampled = downsampleBuffer(channel, audioContext?.sampleRate ?? 48000, options.sampleRateHz ?? 16000);
+        const pcm = floatTo16BitPCM(downsampled);
+        options.onLevel?.(getPcmLevel(pcm));
+        options.onAudio(pcm);
+      };
+      sourceNode.connect(processorNode);
+      processorNode.connect(audioContext.destination);
+    } catch (error) {
+      stop();
+      throw error;
+    }
   };
   const stop = () => {
     processorNode?.disconnect();
@@ -502,6 +507,7 @@ var noopUnsubscribe = () => noop;
 var NOOP_CONNECTION = {
   callControl: noop,
   close: noop,
+  disconnect: noop,
   endTurn: noop,
   send: noop,
   sendAudio: noop,
@@ -513,7 +519,7 @@ var NOOP_CONNECTION = {
   start: () => {}
 };
 var createSessionId = () => crypto.randomUUID();
-var buildWsUrl = (path, sessionId, scenarioId) => {
+var buildWsUrl = (path, sessionId, scenarioId, query) => {
   const { hostname, port, protocol } = window.location;
   const wsProtocol = protocol === "https:" ? "wss:" : "ws:";
   const portSuffix = port ? `:${port}` : "";
@@ -522,6 +528,11 @@ var buildWsUrl = (path, sessionId, scenarioId) => {
   if (scenarioId) {
     url.searchParams.set(DEFAULT_SCENARIO_QUERY_PARAM, scenarioId);
   }
+  for (const [key, value] of Object.entries(query)) {
+    if (key !== "sessionId" && key !== DEFAULT_SCENARIO_QUERY_PARAM && value) {
+      url.searchParams.set(key, value);
+    }
+  }
   return url.toString();
 };
 var isVoiceServerMessage = (value) => {
@@ -631,7 +642,7 @@ var createVoiceConnection = (path, options = {}) => {
     }, delayMs);
   };
   const connect = () => {
-    const ws = new WebSocket(buildWsUrl(path, state.sessionId, state.scenarioId));
+    const ws = new WebSocket(buildWsUrl(path, state.sessionId, state.scenarioId, options.query ?? {}));
     ws.binaryType = "arraybuffer";
     ws.onopen = () => {
       const wasReconnecting = state.reconnectAttempts > 0;
@@ -727,7 +738,19 @@ var createVoiceConnection = (path, options = {}) => {
       type: "call_control"
     });
   };
-  const close = () => {
+  const close = (reason = "client-close") => {
+    clearTimers();
+    if (state.ws) {
+      if (state.ws.readyState === WS_OPEN) {
+        state.ws.send(JSON.stringify({ reason, type: "close" }));
+      }
+      state.ws.close(WS_NORMAL_CLOSURE);
+      state.ws = null;
+    }
+    state.isConnected = false;
+    listeners.clear();
+  };
+  const disconnect = () => {
     clearTimers();
     if (state.ws) {
       state.ws.close(WS_NORMAL_CLOSURE);
@@ -751,6 +774,7 @@ var createVoiceConnection = (path, options = {}) => {
   return {
     callControl,
     close,
+    disconnect,
     endTurn,
     send,
     sendAudio,
@@ -1031,10 +1055,17 @@ var createVoiceStream = (path, options = {}) => {
     callControl(message) {
       connection.callControl(message);
     },
-    close() {
+    close(reason) {
       unsubscribeConnection();
       browserMediaReporter?.close();
-      connection.close();
+      connection.close(reason);
+      store.dispatch({ type: "disconnected" });
+      notify();
+    },
+    disconnect() {
+      unsubscribeConnection();
+      browserMediaReporter?.close();
+      connection.disconnect();
       store.dispatch({ type: "disconnected" });
       notify();
     },
@@ -1479,6 +1510,7 @@ var createVoiceController = (path, options = {}) => {
       };
       notify();
     } catch (error) {
+      capture?.stop();
       capture = null;
       state = {
         ...state,
@@ -1489,13 +1521,19 @@ var createVoiceController = (path, options = {}) => {
       throw error;
     }
   };
-  const close = () => {
+  const close = (reason) => {
+    unsubscribeStream();
+    stopRecording();
+    stream.close(reason);
+  };
+  const disconnect = () => {
     unsubscribeStream();
     stopRecording();
-    stream.close();
+    stream.disconnect();
   };
   return {
     close,
+    disconnect,
     startRecording,
     stopRecording,
     get assistantAudio() {

package/dist/client/index.js

@@ -115,6 +115,7 @@ var noopUnsubscribe = () => noop;
 var NOOP_CONNECTION = {
   callControl: noop,
   close: noop,
+  disconnect: noop,
   endTurn: noop,
   send: noop,
   sendAudio: noop,
@@ -126,7 +127,7 @@ var NOOP_CONNECTION = {
   start: () => {}
 };
 var createSessionId = () => crypto.randomUUID();
-var buildWsUrl = (path, sessionId, scenarioId) => {
+var buildWsUrl = (path, sessionId, scenarioId, query) => {
   const { hostname, port, protocol } = window.location;
   const wsProtocol = protocol === "https:" ? "wss:" : "ws:";
   const portSuffix = port ? `:${port}` : "";
@@ -135,6 +136,11 @@ var buildWsUrl = (path, sessionId, scenarioId) => {
   if (scenarioId) {
     url.searchParams.set(DEFAULT_SCENARIO_QUERY_PARAM, scenarioId);
   }
+  for (const [key, value] of Object.entries(query)) {
+    if (key !== "sessionId" && key !== DEFAULT_SCENARIO_QUERY_PARAM && value) {
+      url.searchParams.set(key, value);
+    }
+  }
   return url.toString();
 };
 var isVoiceServerMessage = (value) => {
@@ -244,7 +250,7 @@ var createVoiceConnection = (path, options = {}) => {
     }, delayMs);
   };
   const connect = () => {
-    const ws = new WebSocket(buildWsUrl(path, state.sessionId, state.scenarioId));
+    const ws = new WebSocket(buildWsUrl(path, state.sessionId, state.scenarioId, options.query ?? {}));
     ws.binaryType = "arraybuffer";
     ws.onopen = () => {
       const wasReconnecting = state.reconnectAttempts > 0;
@@ -340,7 +346,19 @@ var createVoiceConnection = (path, options = {}) => {
       type: "call_control"
     });
   };
-  const close = () => {
+  const close = (reason = "client-close") => {
+    clearTimers();
+    if (state.ws) {
+      if (state.ws.readyState === WS_OPEN) {
+        state.ws.send(JSON.stringify({ reason, type: "close" }));
+      }
+      state.ws.close(WS_NORMAL_CLOSURE);
+      state.ws = null;
+    }
+    state.isConnected = false;
+    listeners.clear();
+  };
+  const disconnect = () => {
     clearTimers();
     if (state.ws) {
       state.ws.close(WS_NORMAL_CLOSURE);
@@ -364,6 +382,7 @@ var createVoiceConnection = (path, options = {}) => {
   return {
     callControl,
     close,
+    disconnect,
     endTurn,
     send,
     sendAudio,
@@ -1563,10 +1582,17 @@ var createVoiceStream = (path, options = {}) => {
     callControl(message) {
       connection.callControl(message);
     },
-    close() {
+    close(reason) {
+      unsubscribeConnection();
+      browserMediaReporter?.close();
+      connection.close(reason);
+      store.dispatch({ type: "disconnected" });
+      notify();
+    },
+    disconnect() {
       unsubscribeConnection();
       browserMediaReporter?.close();
-      connection.close();
+      connection.disconnect();
       store.dispatch({ type: "disconnected" });
       notify();
     },
@@ -1730,29 +1756,34 @@ var createMicrophoneCapture = (options) => {
     if (!AudioContextCtor) {
       throw new Error("Browser microphone capture requires AudioContext support.");
     }
-    mediaStream = options.stream ?? await navigator.mediaDevices.getUserMedia({
-      audio: {
-        autoGainControl: true,
-        channelCount: options.channelCount ?? 1,
-        echoCancellation: true,
-        noiseSuppression: true
+    try {
+      mediaStream = options.stream ?? await navigator.mediaDevices.getUserMedia({
+        audio: {
+          autoGainControl: true,
+          channelCount: options.channelCount ?? 1,
+          echoCancellation: true,
+          noiseSuppression: true
+        }
+      });
+      audioContext = new AudioContextCtor;
+      if (audioContext.state === "suspended") {
+        await audioContext.resume();
       }
-    });
-    audioContext = new AudioContextCtor;
-    if (audioContext.state === "suspended") {
-      await audioContext.resume();
-    }
-    sourceNode = audioContext.createMediaStreamSource(mediaStream);
-    processorNode = audioContext.createScriptProcessor(4096, 1, 1);
-    processorNode.onaudioprocess = (event) => {
-      const channel = event.inputBuffer.getChannelData(0);
-      const downsampled = downsampleBuffer(channel, audioContext?.sampleRate ?? 48000, options.sampleRateHz ?? 16000);
-      const pcm = floatTo16BitPCM(downsampled);
-      options.onLevel?.(getPcmLevel(pcm));
-      options.onAudio(pcm);
-    };
-    sourceNode.connect(processorNode);
-    processorNode.connect(audioContext.destination);
+      sourceNode = audioContext.createMediaStreamSource(mediaStream);
+      processorNode = audioContext.createScriptProcessor(4096, 1, 1);
+      processorNode.onaudioprocess = (event) => {
+        const channel = event.inputBuffer.getChannelData(0);
+        const downsampled = downsampleBuffer(channel, audioContext?.sampleRate ?? 48000, options.sampleRateHz ?? 16000);
+        const pcm = floatTo16BitPCM(downsampled);
+        options.onLevel?.(getPcmLevel(pcm));
+        options.onAudio(pcm);
+      };
+      sourceNode.connect(processorNode);
+      processorNode.connect(audioContext.destination);
+    } catch (error) {
+      stop();
+      throw error;
+    }
   };
   const stop = () => {
     processorNode?.disconnect();
@@ -2195,6 +2226,7 @@ var createVoiceController = (path, options = {}) => {
       };
       notify();
     } catch (error) {
+      capture?.stop();
       capture = null;
       state = {
         ...state,
@@ -2205,13 +2237,19 @@ var createVoiceController = (path, options = {}) => {
       throw error;
     }
   };
-  const close = () => {
+  const close = (reason) => {
+    unsubscribeStream();
+    stopRecording();
+    stream.close(reason);
+  };
+  const disconnect = () => {
     unsubscribeStream();
     stopRecording();
-    stream.close();
+    stream.disconnect();
   };
   return {
     close,
+    disconnect,
     startRecording,
     stopRecording,
     get assistantAudio() {

package/dist/core/types.d.ts

@@ -871,6 +871,19 @@ export type VoicePluginConfig<TContext = unknown, TSession extends VoiceSessionR
     };
     costTelemetry?: VoiceCostTelemetryConfig<TContext, TSession, TResult>;
     path: string;
+    /**
+     * Authorize a WebSocket before the plugin supersedes an existing session or
+     * creates provider-backed resources. Return false to close it with 4401.
+     * Applications can validate a short-lived admission credential from `query`
+     * without coupling Voice to a particular authentication implementation.
+     */
+    authorizeConnection?: (input: {
+        context: TContext;
+        path: string;
+        query: Record<string, unknown>;
+        scenarioId?: string;
+        sessionId: string;
+    }) => boolean | Promise<boolean>;
     greeting?: string | ((input: {
         session: TSession;
     }) => string | Promise<string>);
@@ -1331,6 +1344,8 @@ export type VoiceConnectionOptions = {
      *  the caller losing the call. */
     reconnectMaxDelayMs?: number;
     pingInterval?: number;
+    /** Additional query values sent on every initial or reconnecting socket. */
+    query?: Record<string, string>;
     sessionId?: string;
 };
 export type VoiceBrowserMediaReportPayload = {
@@ -1502,7 +1517,9 @@ export type VoiceStreamState<TResult = unknown> = {
 export type VoiceStream<TResult = unknown> = {
     call: VoiceCallLifecycleState | null;
     callControl: (message: Omit<VoiceClientCallControlMessage, "type">) => void;
-    close: () => void;
+    close: (reason?: string) => void;
+    /** Release the transport without terminally completing the server session. */
+    disconnect: () => void;
     start: (input?: {
         scenarioId?: string;
         sessionId?: string;
@@ -1597,7 +1614,9 @@ export type VoiceController<TResult = unknown> = {
     bindHTMX: (options: VoiceHTMXBindingOptions) => () => void;
     call: VoiceCallLifecycleState | null;
     callControl: (message: Omit<VoiceClientCallControlMessage, "type">) => void;
-    close: () => void;
+    close: (reason?: string) => void;
+    /** Release the transport while leaving the server session resumable. */
+    disconnect: () => void;
     endTurn: () => void;
     start: (input?: {
         scenarioId?: string;