@openclaw/voice-call 2026.5.3 → 2026.5.4-beta.2

package/dist/index.js

@@ -1,7 +1,7 @@
 import { definePluginEntry, sleep } from "./runtime-api.js";
 import "./api.js";
 import { i as resolveVoiceCallConfig, s as validateProviderConfig } from "./config-7w04YpHh.js";
-import { a as setupTailscaleExposureRoute, i as getTailscaleSelfInfo, n as resolveWebhookExposureStatus, r as cleanupTailscaleExposureRoute, s as resolveUserPath, t as createVoiceCallRuntime } from "./runtime-entry-88ytYAQa.js";
+import { a as setupTailscaleExposureRoute, i as getTailscaleSelfInfo, n as resolveWebhookExposureStatus, r as cleanupTailscaleExposureRoute, s as resolveUserPath, t as createVoiceCallRuntime } from "./runtime-entry-DFzuGKLG.js";
 import { i as parseVoiceCallPluginConfig, r as normalizeVoiceCallLegacyConfigInput, t as formatVoiceCallLegacyConfigWarnings } from "./config-compat-B0me39_4.js";
 import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { ErrorCodes, callGatewayFromCli, errorShape } from "openclaw/plugin-sdk/gateway-runtime";
@@ -610,6 +610,8 @@ function createVoiceCallContinueOperationStore(params) {
 }
 //#endregion
 //#region extensions/voice-call/index.ts
+const VOICE_CALL_WRITE_METHOD_SCOPE = { scope: "operator.write" };
+const VOICE_CALL_READ_METHOD_SCOPE = { scope: "operator.read" };
 const voiceCallConfigSchema = {
 	parse(value) {
 		const normalized = normalizeVoiceCallLegacyConfigInput(value);
@@ -988,7 +990,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_WRITE_METHOD_SCOPE);
 		api.registerGatewayMethod("voicecall.continue", async ({ params, respond }) => {
 			try {
 				await respondToCallMessageAction({
@@ -1001,7 +1003,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_WRITE_METHOD_SCOPE);
 		api.registerGatewayMethod("voicecall.continue.start", async ({ params, respond }) => {
 			try {
 				const request = await resolveCallMessageRequest(params);
@@ -1013,7 +1015,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_WRITE_METHOD_SCOPE);
 		api.registerGatewayMethod("voicecall.continue.result", async ({ params, respond }) => {
 			try {
 				const operationId = normalizeOptionalString(params?.operationId) ?? "";
@@ -1030,7 +1032,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_READ_METHOD_SCOPE);
 		api.registerGatewayMethod("voicecall.speak", async ({ params, respond }) => {
 			try {
 				const request = await resolveCallMessageRequest(params);
@@ -1039,10 +1041,18 @@ var voice_call_default = definePluginEntry({
 					return;
 				}
 				if (request.rt.config.realtime.enabled) {
-					if (request.rt.webhookServer.speakRealtime(request.callId, request.message).success) {
+					const realtimeResult = request.rt.webhookServer.speakRealtime(request.callId, request.message);
+					if (realtimeResult.success) {
 						respond(true, { success: true });
 						return;
 					}
+					if (params?.allowTwimlFallback === false) {
+						respond(true, {
+							success: false,
+							error: realtimeResult.error ?? "Realtime bridge is not active"
+						});
+						return;
+					}
 				}
 				const result = await request.rt.manager.speak(request.callId, request.message);
 				if (!result.success) {
@@ -1053,7 +1063,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_WRITE_METHOD_SCOPE);
 		api.registerGatewayMethod("voicecall.dtmf", async ({ params, respond }) => {
 			try {
 				const callId = normalizeOptionalString(params?.callId) ?? "";
@@ -1071,7 +1081,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_WRITE_METHOD_SCOPE);
 		api.registerGatewayMethod("voicecall.end", async ({ params, respond }) => {
 			try {
 				const callId = normalizeOptionalString(params?.callId) ?? "";
@@ -1088,7 +1098,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_WRITE_METHOD_SCOPE);
 		api.registerGatewayMethod("voicecall.status", async ({ params, respond }) => {
 			try {
 				const raw = normalizeOptionalString(params?.callId) ?? normalizeOptionalString(params?.sid) ?? "";
@@ -1112,7 +1122,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_READ_METHOD_SCOPE);
 		api.registerGatewayMethod("voicecall.start", async ({ params, respond }) => {
 			try {
 				const to = normalizeOptionalString(params?.to) ?? "";
@@ -1122,10 +1132,9 @@ var voice_call_default = definePluginEntry({
 					respondError(respond, "to required", ErrorCodes.INVALID_REQUEST);
 					return;
 				}
-				const rt = await ensureRuntime();
 				const mode = params?.mode === "notify" || params?.mode === "conversation" ? params.mode : void 0;
 				await initiateCallAndRespond({
-					rt,
+					rt: await ensureRuntime(),
 					respond,
 					to,
 					message: message || void 0,
@@ -1135,7 +1144,7 @@ var voice_call_default = definePluginEntry({
 			} catch (err) {
 				sendError(respond, err);
 			}
-		});
+		}, VOICE_CALL_WRITE_METHOD_SCOPE);
 		api.registerTool({
 			name: "voice_call",
 			label: "Voice Call",

package/dist/{realtime-handler-B63CIDP2.js → realtime-handler-C-SaPrny.js}

@@ -2,10 +2,155 @@ import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME, buildRealtimeVoiceAgentConsultWorkingResponse, createRealtimeVoiceBridgeSession } from "openclaw/plugin-sdk/realtime-voice";
 import { randomUUID } from "node:crypto";
 import WebSocket, { WebSocketServer } from "ws";
+//#region extensions/voice-call/src/webhook/realtime-audio-pacer.ts
+const TELEPHONY_SAMPLE_RATE = 8e3;
+const TELEPHONY_CHUNK_BYTES = 160;
+const TELEPHONY_CHUNK_MS = 20;
+const DEFAULT_SPEECH_RMS_THRESHOLD = .02;
+const DEFAULT_REQUIRED_LOUD_CHUNKS = 2;
+const DEFAULT_REQUIRED_QUIET_CHUNKS = 10;
+const DEFAULT_MAX_QUEUED_AUDIO_BYTES = TELEPHONY_SAMPLE_RATE * 120;
+const PCM16_MAX_AMPLITUDE = 32768;
+const MULAW_LINEAR_SAMPLES = new Int16Array(256);
+for (let i = 0; i < MULAW_LINEAR_SAMPLES.length; i += 1) MULAW_LINEAR_SAMPLES[i] = decodeMulawSample(i);
+var RealtimeTwilioAudioPacer = class {
+	constructor(params) {
+		this.params = params;
+		this.queue = [];
+		this.timer = null;
+		this.queuedAudioBytes = 0;
+		this.closed = false;
+	}
+	sendAudio(muLaw) {
+		if (this.closed || muLaw.length === 0) return;
+		const maxQueuedAudioBytes = this.params.maxQueuedAudioBytes ?? DEFAULT_MAX_QUEUED_AUDIO_BYTES;
+		for (let offset = 0; offset < muLaw.length; offset += TELEPHONY_CHUNK_BYTES) {
+			const chunk = Buffer.from(muLaw.subarray(offset, offset + TELEPHONY_CHUNK_BYTES));
+			if (this.queuedAudioBytes + chunk.length > maxQueuedAudioBytes) {
+				this.failBackpressure();
+				return;
+			}
+			this.queue.push({
+				type: "audio",
+				chunk,
+				durationMs: Math.max(1, Math.round(chunk.length / TELEPHONY_SAMPLE_RATE * 1e3))
+			});
+			this.queuedAudioBytes += chunk.length;
+		}
+		this.ensurePump();
+	}
+	sendMark(name) {
+		if (this.closed || !name) return;
+		this.queue.push({
+			type: "mark",
+			name
+		});
+		this.ensurePump();
+	}
+	clearAudio() {
+		if (this.closed) return;
+		this.clearTimer();
+		this.queue = [];
+		this.queuedAudioBytes = 0;
+		this.params.sendJson({
+			event: "clear",
+			streamSid: this.params.streamSid
+		});
+	}
+	close() {
+		this.closed = true;
+		this.clearTimer();
+		this.queue = [];
+		this.queuedAudioBytes = 0;
+	}
+	clearTimer() {
+		if (!this.timer) return;
+		clearTimeout(this.timer);
+		this.timer = null;
+	}
+	ensurePump() {
+		if (!this.timer) this.pump();
+	}
+	failBackpressure() {
+		this.close();
+		this.params.onBackpressure?.();
+	}
+	pump() {
+		this.timer = null;
+		if (this.closed) return;
+		const item = this.queue.shift();
+		if (!item) return;
+		let delayMs = 0;
+		let sent = true;
+		if (item.type === "audio") {
+			this.queuedAudioBytes = Math.max(0, this.queuedAudioBytes - item.chunk.length);
+			sent = this.params.sendJson({
+				event: "media",
+				streamSid: this.params.streamSid,
+				media: { payload: item.chunk.toString("base64") }
+			});
+			delayMs = item.durationMs || TELEPHONY_CHUNK_MS;
+		} else sent = this.params.sendJson({
+			event: "mark",
+			streamSid: this.params.streamSid,
+			mark: { name: item.name }
+		});
+		if (!sent) {
+			this.queue = [];
+			this.queuedAudioBytes = 0;
+			return;
+		}
+		if (this.queue.length > 0) this.timer = setTimeout(() => this.pump(), delayMs);
+	}
+};
+function calculateMulawRms(muLaw) {
+	if (muLaw.length === 0) return 0;
+	let sum = 0;
+	for (let i = 0; i < muLaw.length; i += 1) {
+		const normalized = (MULAW_LINEAR_SAMPLES[muLaw[i] ?? 0] ?? 0) / PCM16_MAX_AMPLITUDE;
+		sum += normalized * normalized;
+	}
+	return Math.sqrt(sum / muLaw.length);
+}
+var RealtimeMulawSpeechStartDetector = class {
+	constructor(params = {}) {
+		this.params = params;
+		this.loudChunks = 0;
+		this.quietChunks = DEFAULT_REQUIRED_QUIET_CHUNKS;
+		this.speaking = false;
+	}
+	accept(muLaw) {
+		if (calculateMulawRms(muLaw) >= (this.params.rmsThreshold ?? DEFAULT_SPEECH_RMS_THRESHOLD)) {
+			this.quietChunks = 0;
+			this.loudChunks += 1;
+			const requiredLoudChunks = this.params.requiredLoudChunks ?? DEFAULT_REQUIRED_LOUD_CHUNKS;
+			if (!this.speaking && this.loudChunks >= requiredLoudChunks) {
+				this.speaking = true;
+				return true;
+			}
+			return false;
+		}
+		this.loudChunks = 0;
+		this.quietChunks += 1;
+		const requiredQuietChunks = this.params.requiredQuietChunks ?? DEFAULT_REQUIRED_QUIET_CHUNKS;
+		if (this.quietChunks >= requiredQuietChunks) this.speaking = false;
+		return false;
+	}
+};
+function decodeMulawSample(value) {
+	const muLaw = ~value & 255;
+	const sign = muLaw & 128;
+	const exponent = muLaw >> 4 & 7;
+	let sample = ((muLaw & 15) << 3) + 132 << exponent;
+	sample -= 132;
+	return sign ? -sample : sample;
+}
+//#endregion
 //#region extensions/voice-call/src/webhook/realtime-handler.ts
 const STREAM_TOKEN_TTL_MS = 3e4;
 const DEFAULT_HOST = "localhost:8443";
 const MAX_REALTIME_MESSAGE_BYTES = 256 * 1024;
+const MAX_REALTIME_WS_BUFFERED_BYTES = 1024 * 1024;
 function normalizePath(pathname) {
 	const trimmed = pathname.trim();
 	if (!trimmed) return "/";
@@ -99,7 +244,8 @@ var RealtimeCallHandler = class {
 					if (!bridge) return;
 					const mediaData = typeof msg.media === "object" && msg.media !== null ? msg.media : void 0;
 					if (msg.event === "media" && typeof mediaData?.payload === "string") {
-						bridge.sendAudio(Buffer.from(mediaData.payload, "base64"));
+						const audio = Buffer.from(mediaData.payload, "base64");
+						bridge.sendAudio(audio);
 						if (typeof mediaData.timestamp === "number") bridge.setMediaTimestamp(mediaData.timestamp);
 						else if (typeof mediaData.timestamp === "string") bridge.setMediaTimestamp(Number.parseInt(mediaData.timestamp, 10));
 						return;
@@ -174,7 +320,28 @@ var RealtimeCallHandler = class {
 			callEndEmitted = true;
 			this.endCallInManager(callSid, callId, reason);
 		};
-		const bridge = createRealtimeVoiceBridgeSession({
+		const sendJson = (message) => {
+			if (ws.readyState !== WebSocket.OPEN) return false;
+			if (ws.bufferedAmount > MAX_REALTIME_WS_BUFFERED_BYTES) {
+				ws.close(1013, "Backpressure: send buffer exceeded");
+				return false;
+			}
+			ws.send(JSON.stringify(message));
+			if (ws.bufferedAmount > MAX_REALTIME_WS_BUFFERED_BYTES) {
+				ws.close(1013, "Backpressure: send buffer exceeded");
+				return false;
+			}
+			return true;
+		};
+		const audioPacer = new RealtimeTwilioAudioPacer({
+			streamSid,
+			sendJson,
+			onBackpressure: () => {
+				if (ws.readyState === WebSocket.OPEN) ws.close(1013, "Backpressure: paced audio queue exceeded");
+			}
+		});
+		const speechDetector = new RealtimeMulawSpeechStartDetector();
+		const session = createRealtimeVoiceBridgeSession({
 			provider: this.realtimeProvider,
 			providerConfig: this.providerConfig,
 			instructions: this.config.instructions,
@@ -184,24 +351,13 @@ var RealtimeCallHandler = class {
 			audioSink: {
 				isOpen: () => ws.readyState === WebSocket.OPEN,
 				sendAudio: (muLaw) => {
-					ws.send(JSON.stringify({
-						event: "media",
-						streamSid,
-						media: { payload: muLaw.toString("base64") }
-					}));
+					audioPacer.sendAudio(muLaw);
 				},
 				clearAudio: () => {
-					ws.send(JSON.stringify({
-						event: "clear",
-						streamSid
-					}));
+					audioPacer.clearAudio();
 				},
 				sendMark: (markName) => {
-					ws.send(JSON.stringify({
-						event: "mark",
-						streamSid,
-						mark: { name: markName }
-					}));
+					audioPacer.sendMark(markName);
 				}
 			},
 			onTranscript: (role, text, isFinal) => {
@@ -242,7 +398,10 @@ var RealtimeCallHandler = class {
 				this.activeBridgesByCallId.delete(callId);
 				this.activeBridgesByCallId.delete(callSid);
 				this.partialUserTranscriptsByCallId.delete(callId);
-				if (reason !== "error") return;
+				if (reason !== "error") {
+					emitCallEnd("completed");
+					return;
+				}
 				emitCallEnd("error");
 				if (ws.readyState === WebSocket.OPEN) ws.close(1011, "Bridge disconnected");
 				this.provider.hangupCall({
@@ -254,22 +413,28 @@ var RealtimeCallHandler = class {
 				});
 			}
 		});
-		this.activeBridgesByCallId.set(callId, bridge);
-		this.activeBridgesByCallId.set(callSid, bridge);
-		const closeBridge = bridge.close.bind(bridge);
-		bridge.close = () => {
+		this.activeBridgesByCallId.set(callId, session);
+		this.activeBridgesByCallId.set(callSid, session);
+		const sendAudioToSession = session.sendAudio.bind(session);
+		session.sendAudio = (audio) => {
+			if (speechDetector.accept(audio)) audioPacer.clearAudio();
+			sendAudioToSession(audio);
+		};
+		const closeSession = session.close.bind(session);
+		session.close = () => {
 			this.activeBridgesByCallId.delete(callId);
 			this.activeBridgesByCallId.delete(callSid);
 			this.partialUserTranscriptsByCallId.delete(callId);
-			closeBridge();
+			audioPacer.close();
+			closeSession();
 		};
-		bridge.connect().catch((error) => {
+		session.connect().catch((error) => {
 			console.error("[voice-call] Failed to connect realtime bridge:", error);
-			bridge.close();
+			session.close();
 			emitCallEnd("error");
 			ws.close(1011, "Failed to connect");
 		});
-		return bridge;
+		return session;
 	}
 	registerCallInManager(callSid, callerMeta = {}) {
 		const baseFields = {

package/dist/{runtime-entry-88ytYAQa.js → runtime-entry-DFzuGKLG.js}

@@ -2861,7 +2861,7 @@ function loadRealtimeVoiceRuntime() {
 	return realtimeVoiceRuntimePromise;
 }
 function loadRealtimeHandler() {
-	realtimeHandlerPromise ??= import("./realtime-handler-B63CIDP2.js");
+	realtimeHandlerPromise ??= import("./realtime-handler-C-SaPrny.js");
 	return realtimeHandlerPromise;
 }
 function resolveVoiceCallConsultSessionKey(call) {

package/dist/runtime-entry.js

@@ -1,2 +1,2 @@
-import { t as createVoiceCallRuntime } from "./runtime-entry-88ytYAQa.js";
+import { t as createVoiceCallRuntime } from "./runtime-entry-DFzuGKLG.js";
 export { createVoiceCallRuntime };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@openclaw/voice-call",
-  "version": "2026.5.3",
+  "version": "2026.5.4-beta.2",
   "description": "OpenClaw voice-call plugin",
   "repository": {
     "type": "git",
@@ -17,7 +17,7 @@
     "openclaw": "workspace:*"
   },
   "peerDependencies": {
-    "openclaw": ">=2026.5.3"
+    "openclaw": ">=2026.5.4-beta.2"
   },
   "peerDependenciesMeta": {
     "openclaw": {
@@ -34,10 +34,10 @@
       "minHostVersion": ">=2026.4.10"
     },
     "compat": {
-      "pluginApi": ">=2026.5.3"
+      "pluginApi": ">=2026.5.4-beta.2"
     },
     "build": {
-      "openclawVersion": "2026.5.3"
+      "openclawVersion": "2026.5.4-beta.2"
     },
     "release": {
       "publishToClawHub": true,