@openclaw/voice-call 2026.9.7 → 2026.10.1-beta.1

package/dist/.setup/{realtime-handler-BgDPcPZt.mjs → realtime-handler-Do7Sz7Gh.mjs}

@@ -1,5 +1,5 @@
-import { d as REALTIME_VOICE_END_CALL_TOOL_NAME, s as resolveVoiceCallPublicPathPrefix } from "./config-Ds7xoZbI.mjs";
-import { c as canonicalizeVoiceCallMediaBase64, o as WebSocket, s as WebSocketServer } from "./runtime-entry-BNOQRXii.mjs";
+import { d as REALTIME_VOICE_END_CALL_TOOL_NAME, s as resolveVoiceCallPublicPathPrefix } from "./config-CX9J-053.mjs";
+import { o as canonicalizeVoiceCallMediaBase64 } from "./runtime-entry-CbSKxSKk.mjs";
 import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { asNullableRecord, asOptionalObjectRecord, asOptionalRecord, normalizeOptionalString } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { createSubsystemLogger } from "openclaw/plugin-sdk/runtime-env";
@@ -7,8 +7,10 @@ import { REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME, REALTIME_VOICE_AUDIO_FORMAT_G71
 import { normalizeWebhookPath } from "openclaw/plugin-sdk/webhook-ingress";
 import { isFutureDateTimestampMs, resolveExpiresAtMsFromDurationMs } from "openclaw/plugin-sdk/number-runtime";
 import { randomUUID } from "node:crypto";
-import { sliceUtf16Safe, truncateUtf16Safe } from "openclaw/plugin-sdk/text-utility-runtime";
+import { safeParseJson, sliceUtf16Safe } from "openclaw/plugin-sdk/text-utility-runtime";
 import "node:http";
+import { WebSocket, WebSocketServer, rejectWebSocketUpgrade } from "openclaw/plugin-sdk/websocket-runtime";
+import { createDeferred } from "openclaw/plugin-sdk/extension-shared";
 //#region extensions/voice-call/src/webhook/realtime-audio-pacer.ts
 const TELEPHONY_SAMPLE_RATE = 8e3;
 const TELEPHONY_CHUNK_BYTES = 160;
@@ -17,7 +19,6 @@ const DEFAULT_MAX_QUEUED_AUDIO_BYTES = TELEPHONY_SAMPLE_RATE * 120;
 const QUEUE_COMPACT_HEAD_THRESHOLD = 256;
 const MAX_PLAYBACK_SEGMENTS = 128;
 const MAX_PENDING_MARK_BOUNDARIES = 64;
-/** Paces outgoing mulaw audio frames at telephony cadence. */
 var RealtimeAudioPacer = class {
 	constructor(params) {
 		this.params = params;
@@ -103,11 +104,9 @@ var RealtimeAudioPacer = class {
 		while (this.playbackSegments.length > 0) {
 			const head = this.playbackSegments[0];
 			if (!head || head.sentMs < head.totalMs || head.lastSentEndMs === void 0 || head.lastSentEndMs > boundary.sentMs) break;
-			const retired = this.playbackSegments.shift();
-			if (retired) {
-				this.retiredAudioMs += retired.sentMs;
-				if (retired.itemId !== void 0) this.retiredItemOffsets.set(retired.itemId, (this.retiredItemOffsets.get(retired.itemId) ?? 0) + retired.sentMs);
-			}
+			this.playbackSegments.shift();
+			this.retiredAudioMs += head.sentMs;
+			if (head.itemId !== void 0) this.retiredItemOffsets.set(head.itemId, (this.retiredItemOffsets.get(head.itemId) ?? 0) + head.sentMs);
 		}
 		this.confirmedPlayedMs = Math.max(this.confirmedPlayedMs, Math.min(boundary.sentMs, this.sentAudioMs));
 	}
@@ -118,14 +117,13 @@ var RealtimeAudioPacer = class {
 		this.clearTimer();
 		this.resetQueue();
 		this.resetPlaybackState();
-		this.params.send(this.params.serializer.clear());
+		this.params.send(this.params.serializer.serializeClear());
 		return clearedAudioBytes;
 	}
 	/** True while queued audio or a paced send timer can still reach the telephony stream. */
 	hasPendingAudio() {
 		return !this.closed && (this.queuedAudioBytes > 0 || this.timer !== null);
 	}
-	/** Stop sending and discard queued frames. */
 	close() {
 		this.closed = true;
 		this.clearTimer();
@@ -158,17 +156,14 @@ var RealtimeAudioPacer = class {
 		this.streamClockMs = null;
 		this.params.onPlaybackReset?.();
 	}
-	/** Clear the scheduled pump timer. */
 	clearTimer() {
 		if (!this.timer) return;
 		clearTimeout(this.timer);
 		this.timer = null;
 	}
-	/** Start the pump when queued work exists and no timer is active. */
 	ensurePump() {
 		if (!this.timer) this.pump();
 	}
-	/** Close the pacer and notify the caller about queued-audio backpressure. */
 	failBackpressure() {
 		this.close();
 		this.params.onBackpressure?.();
@@ -220,7 +215,7 @@ var RealtimeAudioPacer = class {
 	}
 	sendAudioItem(item) {
 		this.queuedAudioBytes = Math.max(0, this.queuedAudioBytes - item.chunk.length);
-		const sent = this.params.send(this.params.serializer.media(item.chunk.toString("base64")));
+		const sent = this.params.send(this.params.serializer.serializeMedia(item.chunk.toString("base64")));
 		if (sent) {
 			item.segment.sentMs += item.durationMs;
 			this.sentAudioMs += item.durationMs;
@@ -237,7 +232,7 @@ var RealtimeAudioPacer = class {
 	}
 	/** Send a queued mark frame and bind it to the playback prefix before it. */
 	sendMarkItem(item) {
-		const sent = this.params.send(this.params.serializer.mark(item.name));
+		const sent = this.params.send(this.params.serializer.serializeMark(item.name));
 		if (sent) {
 			this.markBoundaries.push({
 				name: item.name,
@@ -258,144 +253,77 @@ function parseTimestampMs(value) {
 		return Number.isSafeInteger(parsed) ? parsed : void 0;
 	}
 }
-/** Parse a JSON object frame, returning null for invalid or non-object payloads. */
-function tryParseJson(rawMessage) {
-	try {
-		const parsed = JSON.parse(rawMessage);
-		return asNullableRecord(parsed);
-	} catch {}
-	return null;
-}
-/** Parse a common provider media frame. */
-function parseMediaFrame(msg) {
-	const mediaData = asOptionalObjectRecord(msg.media);
-	const payload = typeof mediaData?.payload === "string" ? mediaData.payload : void 0;
-	const canonicalPayload = payload ? canonicalizeVoiceCallMediaBase64(payload) : void 0;
-	if (!canonicalPayload) return { kind: "ignored" };
-	return {
-		kind: "media",
-		payloadBase64: canonicalPayload,
-		timestampMs: parseTimestampMs(mediaData?.timestamp),
-		track: typeof mediaData?.track === "string" ? mediaData.track : void 0
-	};
-}
-/** Parse a common provider mark frame. */
-function parseMarkFrame(msg) {
-	const markData = asOptionalObjectRecord(msg.mark);
-	return {
-		kind: "mark",
-		name: typeof markData?.name === "string" ? markData.name : void 0
-	};
-}
-/** Parse one provider frame with provider-specific start/error hooks. */
-function parseProviderInboundFrame(rawMessage, parseStartFrame, parseExtraFrame) {
-	const msg = tryParseJson(rawMessage);
-	if (!msg) return { kind: "ignored" };
-	const event = msg.event;
-	switch (event) {
-		case "start": return parseStartFrame(msg) ?? { kind: "ignored" };
-		case "media": return parseMediaFrame(msg);
-		case "mark": return parseMarkFrame(msg);
-		case "stop": return { kind: "stop" };
-		default: return parseExtraFrame?.(event, msg) ?? { kind: "ignored" };
-	}
-}
-/** Serialize a provider media frame. */
-function serializeMediaFrame(payloadBase64, streamSid) {
-	return JSON.stringify({
-		event: "media",
-		streamSid,
-		media: { payload: payloadBase64 }
-	});
-}
-/** Serialize a provider clear frame. */
-function serializeClearFrame(streamSid) {
-	return JSON.stringify({
-		event: "clear",
-		streamSid
-	});
-}
-/** Serialize a provider mark frame. */
-function serializeMarkFrame(name, streamSid) {
-	return JSON.stringify({
-		event: "mark",
-		streamSid,
-		mark: { name }
-	});
-}
-/** Twilio media stream adapter, retaining streamSid for outbound frames. */
-var TwilioStreamFrameAdapter = class {
-	constructor() {
-		this.providerName = "twilio";
-		this.streamSid = "";
+var StreamFrameAdapter = class {
+	constructor(provider) {
+		this.provider = provider;
+		this.streamSid = provider === "twilio" ? "" : void 0;
 	}
-	/** Parse one Twilio websocket message into a normalized frame. */
 	parseInbound(rawMessage) {
-		return parseProviderInboundFrame(rawMessage, (msg) => {
-			const startData = asOptionalObjectRecord(msg.start);
-			const streamSid = typeof startData?.streamSid === "string" ? startData.streamSid : "";
-			const callSid = typeof startData?.callSid === "string" ? startData.callSid : "";
-			if (!streamSid || !callSid) return;
-			this.streamSid = streamSid;
-			return {
-				kind: "start",
-				streamId: streamSid,
-				providerCallId: callSid
-			};
-		});
+		const msg = asNullableRecord(safeParseJson(rawMessage));
+		if (!msg) return { kind: "ignored" };
+		switch (msg.event) {
+			case "start": {
+				const start = asOptionalObjectRecord(msg.start);
+				const streamId = this.provider === "twilio" ? start?.streamSid : msg.stream_id;
+				const providerCallId = this.provider === "twilio" ? start?.callSid : start?.call_control_id;
+				if (typeof streamId !== "string" || !streamId || typeof providerCallId !== "string" || !providerCallId) return { kind: "ignored" };
+				if (this.provider === "twilio") this.streamSid = streamId;
+				return {
+					kind: "start",
+					streamId,
+					providerCallId
+				};
+			}
+			case "media": {
+				const media = asOptionalObjectRecord(msg.media);
+				const payload = typeof media?.payload === "string" ? media.payload : void 0;
+				const payloadBase64 = payload ? canonicalizeVoiceCallMediaBase64(payload) : void 0;
+				return payloadBase64 ? {
+					kind: "media",
+					payloadBase64,
+					timestampMs: parseTimestampMs(media?.timestamp)
+				} : { kind: "ignored" };
+			}
+			case "mark": {
+				const mark = asOptionalObjectRecord(msg.mark);
+				return {
+					kind: "mark",
+					name: typeof mark?.name === "string" ? mark.name : void 0
+				};
+			}
+			case "stop": return { kind: "stop" };
+			case "error": {
+				if (this.provider !== "telnyx") return { kind: "ignored" };
+				const error = asOptionalObjectRecord(msg.payload);
+				return {
+					kind: "error",
+					code: typeof error?.code === "string" || typeof error?.code === "number" ? String(error.code) : void 0,
+					title: typeof error?.title === "string" ? error.title : void 0,
+					detail: typeof error?.detail === "string" ? error.detail : void 0
+				};
+			}
+			default: return { kind: "ignored" };
+		}
 	}
-	/** Serialize Twilio media with the active streamSid. */
 	serializeMedia(payloadBase64) {
-		return serializeMediaFrame(payloadBase64, this.streamSid);
-	}
-	/** Serialize Twilio clear with the active streamSid. */
-	serializeClear() {
-		return serializeClearFrame(this.streamSid);
-	}
-	/** Serialize Twilio mark with the active streamSid. */
-	serializeMark(name) {
-		return serializeMarkFrame(name, this.streamSid);
-	}
-};
-/** Telnyx media stream adapter. */
-var TelnyxStreamFrameAdapter = class {
-	constructor() {
-		this.providerName = "telnyx";
-	}
-	/** Parse one Telnyx websocket message into a normalized frame. */
-	parseInbound(rawMessage) {
-		return parseProviderInboundFrame(rawMessage, (msg) => {
-			const topLevelStreamId = typeof msg.stream_id === "string" && msg.stream_id ? msg.stream_id : void 0;
-			const startData = asOptionalObjectRecord(msg.start);
-			const providerCallId = typeof startData?.call_control_id === "string" && startData.call_control_id ? startData.call_control_id : void 0;
-			if (!topLevelStreamId || !providerCallId) return;
-			return {
-				kind: "start",
-				streamId: topLevelStreamId,
-				providerCallId
-			};
-		}, (event, msg) => {
-			if (event !== "error") return;
-			const errorData = asOptionalObjectRecord(msg.payload);
-			return {
-				kind: "error",
-				code: typeof errorData?.code === "string" || typeof errorData?.code === "number" ? String(errorData.code) : void 0,
-				title: typeof errorData?.title === "string" ? errorData.title : void 0,
-				detail: typeof errorData?.detail === "string" ? errorData.detail : void 0
-			};
+		return JSON.stringify({
+			event: "media",
+			streamSid: this.streamSid,
+			media: { payload: payloadBase64 }
 		});
 	}
-	/** Serialize Telnyx media. */
-	serializeMedia(payloadBase64) {
-		return serializeMediaFrame(payloadBase64);
-	}
-	/** Serialize Telnyx clear. */
 	serializeClear() {
-		return serializeClearFrame();
+		return JSON.stringify({
+			event: "clear",
+			streamSid: this.streamSid
+		});
 	}
-	/** Serialize Telnyx mark. */
 	serializeMark(name) {
-		return serializeMarkFrame(name);
+		return JSON.stringify({
+			event: "mark",
+			streamSid: this.streamSid,
+			mark: { name }
+		});
 	}
 };
 //#endregion
@@ -430,12 +358,6 @@ function findTextOverlap(base, next) {
 	for (let size = max; size > 0; size -= 1) if (base.slice(-size) === next.slice(0, size)) return size;
 	return 0;
 }
-function shouldInsertTranscriptSpace(base, next) {
-	if (!base || !next) return false;
-	const last = base.at(-1);
-	if (/\s$/.test(base) || last === "(" || last === "[" || last === "{" || last === "\"" || last === "'" || /^[\s,.;:!?)]/.test(next)) return false;
-	return true;
-}
 function appendTranscriptText(base, fragment) {
 	const next = normalizeTranscriptText(fragment);
 	if (!next) return base ?? "";
@@ -447,7 +369,7 @@ function appendTranscriptText(base, fragment) {
 	if (nextLower.startsWith(currentLower)) return next;
 	const overlap = findTextOverlap(currentLower, nextLower);
 	if (overlap >= 6 || overlap >= 3 && next.length <= 12) return `${current}${next.slice(overlap)}`.trim();
-	return `${current}${shouldInsertTranscriptSpace(current, next) ? " " : ""}${next}`.trim();
+	return `${current}${/[([{"']$/.test(current) || /^[,.;:!?)]/.test(next) ? "" : " "}${next}`.trim();
 }
 function resolveFinalTranscriptText(params) {
 	const final = normalizeTranscriptText(params.final);
@@ -474,32 +396,29 @@ function withFallbackConsultQuestion(args, fallback) {
 	const question = fallback?.trim();
 	if (providerQuestion) {
 		if (question && providerQuestion.length <= 40 && question.length >= providerQuestion.length + 8) {
-			const context = normalizeOptionalString(asOptionalRecord(args)?.context);
+			const record = asOptionalRecord(args);
+			const context = normalizeOptionalString(record?.context);
 			const fallbackContext = `Realtime provider supplied a shorter consult question: ${providerQuestion}`;
-			return args && typeof args === "object" && !Array.isArray(args) ? {
-				...args,
+			return {
+				...record,
 				question,
 				context: context ? `${context}\n\n${fallbackContext}` : fallbackContext
-			} : {
-				question,
-				context: fallbackContext
 			};
 		}
 		return args;
 	}
 	if (!question) return args;
-	return args && typeof args === "object" && !Array.isArray(args) ? {
-		...args,
+	return {
+		...asOptionalRecord(args),
 		question
-	} : { question };
+	};
 }
 function buildForcedConsultSpeechPrompt(result) {
-	const trimmed = result.trim();
 	return [
 		"Internal OpenClaw consult result is ready.",
 		"Do not call tools for this internal result.",
 		"Speak the following answer to the caller now, briefly and naturally:",
-		trimmed.length <= FORCED_CONSULT_RESULT_MAX_CHARS ? trimmed : `${truncateUtf16Safe(trimmed, 1784).trimEnd()} [truncated]`
+		result
 	].join("\n");
 }
 async function waitForNativeConsult(state) {
@@ -530,15 +449,6 @@ function appendRecentTalkEventMetadata(metadata, event) {
 		}].slice(-12)
 	};
 }
-function rejectRealtimeUpgrade(socket, status) {
-	const reason = status === 401 ? "Unauthorized" : "Service Unavailable";
-	try {
-		socket.end(`HTTP/1.1 ${status} ${reason}\r\nConnection: close\r\n\r\n`, () => socket.destroy());
-	} catch (error) {
-		socket.destroy();
-		throw error;
-	}
-}
 var RealtimeCallHandler = class {
 	constructor(config, manager, resolveCallRegistration, servePath, streamDisconnectLifecycle, coreConfig) {
 		this.config = config;
@@ -605,16 +515,16 @@ var RealtimeCallHandler = class {
 	handleWebSocketUpgrade(request, socket, head) {
 		socket.once("error", () => socket.destroy());
 		if (this.closing) {
-			rejectRealtimeUpgrade(socket, 503);
+			rejectWebSocketUpgrade(socket, { status: 503 });
 			return;
 		}
 		const token = new URL(request.url ?? "/", "wss://localhost").pathname.split("/").pop() ?? null;
 		const callerMeta = token ? this.consumeStreamToken(token) : null;
 		if (!callerMeta) {
-			rejectRealtimeUpgrade(socket, 401);
+			rejectWebSocketUpgrade(socket, { status: 401 });
 			return;
 		}
-		const adapter = (callerMeta.providerName ?? "twilio") === "telnyx" ? new TelnyxStreamFrameAdapter() : new TwilioStreamFrameAdapter();
+		const adapter = new StreamFrameAdapter(callerMeta.providerName ?? "twilio");
 		new WebSocketServer({
 			noServer: true,
 			maxPayload: MAX_REALTIME_MESSAGE_BYTES
@@ -775,19 +685,13 @@ var RealtimeCallHandler = class {
 		}
 	}
 	issueStreamSession(request = {}) {
-		const token = this.issueStreamToken({
+		const meta = {
 			providerName: request.providerName ?? "twilio",
 			callId: request.callId,
 			from: request.from,
 			to: request.to,
 			direction: request.direction
-		});
-		return {
-			token,
-			streamUrl: `wss://${this.publicOrigin || DEFAULT_HOST}${this.getStreamPathPattern()}/${token}`
 		};
-	}
-	issueStreamToken(meta = {}) {
 		const token = randomUUID();
 		const expiry = resolveExpiresAtMsFromDurationMs(STREAM_TOKEN_TTL_MS, { nowMs: Date.now() });
 		if (expiry !== void 0) {
@@ -807,20 +711,18 @@ var RealtimeCallHandler = class {
 				console.warn(`[voice-call] Realtime stream WebSocket never connected within ${STREAM_TOKEN_TTL_MS / 1e3}s${call}${participants} — the provider could not reach wss://${host}${streamPathPattern}/<token>. Verify the stream path is exposed (tailscale serve/funnel --set-path).`);
 			}, STREAM_TOKEN_TTL_MS).unref?.();
 		}
-		return token;
+		return {
+			token,
+			streamUrl: `wss://${this.publicOrigin || DEFAULT_HOST}${this.getStreamPathPattern()}/${token}`
+		};
 	}
 	consumeStreamToken(token) {
 		const entry = this.pendingStreamTokens.get(token);
 		if (!entry) return null;
 		this.pendingStreamTokens.delete(token);
 		if (!isFutureDateTimestampMs(entry.expiry)) return null;
-		return {
-			from: entry.from,
-			to: entry.to,
-			direction: entry.direction,
-			providerName: entry.providerName,
-			callId: entry.callId
-		};
+		const { expiry: _expiry, ...request } = entry;
+		return request;
 	}
 	async handleCall(streamSid, callSid, ws, callerMeta, adapter) {
 		const preparedCall = await this.prepareCallInManager(callSid, callerMeta);
@@ -992,11 +894,7 @@ var RealtimeCallHandler = class {
 		const audioPacer = new RealtimeAudioPacer({
 			onPlaybackReset: () => pendingMarkAcks.clear(),
 			send: sendString,
-			serializer: {
-				media: (payload) => adapter.serializeMedia(payload),
-				clear: () => adapter.serializeClear(),
-				mark: (name) => adapter.serializeMark(name)
-			},
+			serializer: adapter,
 			onBackpressure: () => {
 				console.warn(`[voice-call] realtime paced audio backpressure callId=${callId} providerCallId=${callSid}`);
 				if (ws.readyState === WebSocket.OPEN) ws.close(1013, "Backpressure: paced audio queue exceeded");
@@ -1509,7 +1407,7 @@ var RealtimeCallHandler = class {
 		const state = this.getUserTranscriptState(callId, owner);
 		if (!state) return;
 		if (state.recentFinalTimer) clearTimeout(state.recentFinalTimer);
-		if (this.userTranscriptStatesByCallId.get(callId) === state) this.userTranscriptStatesByCallId.delete(callId);
+		this.userTranscriptStatesByCallId.delete(callId);
 	}
 	cancelNativeConsult(callId, owner) {
 		const state = this.nativeConsultsInFlightByCallId.get(callId);
@@ -1525,24 +1423,19 @@ var RealtimeCallHandler = class {
 		state.cancel();
 		this.forcedConsultsByCallId.delete(callId);
 	}
-	resetConsultSession(callId, owner) {
-		const session = this.consultSessionsByCallId.get(callId);
-		if (!session || session.owner !== owner) return false;
-		session.coordinator.clearPending();
-		this.cancelForcedConsult(callId, owner);
-		this.cancelNativeConsult(callId, owner);
-		return true;
-	}
 	resetConsultSessionForContinuity(callId, owner) {
 		const session = this.consultSessionsByCallId.get(callId);
-		if (!session || session.owner !== owner) return false;
+		if (!session || session.owner !== owner) return;
 		this.cancelForcedConsult(callId, owner);
 		this.cancelNativeConsult(callId, owner);
 		session.coordinator.clear();
-		return true;
 	}
 	cancelConsultSession(callId, owner) {
-		if (!owner || !this.resetConsultSession(callId, owner)) return;
+		const session = this.consultSessionsByCallId.get(callId);
+		if (!session || session.owner !== owner) return;
+		session.coordinator.clearPending();
+		this.cancelForcedConsult(callId, owner);
+		this.cancelNativeConsult(callId, owner);
 		this.consultSessionsByCallId.delete(callId);
 	}
 	isActiveBridgeOwner(callId, owner) {
@@ -1759,7 +1652,8 @@ var RealtimeCallHandler = class {
 		const hasResultError = (result) => {
 			return result !== null && typeof result === "object" && !Array.isArray(result) && "error" in result;
 		};
-		const emitFinalToolEvent = (result) => {
+		const submitFinalToolResult = async (result) => {
+			if (!delegation) await bridge.submitToolResult(bridgeCallId, result);
 			harness.emit({
 				type: hasResultError(result) ? "tool.error" : "tool.result",
 				turnId,
@@ -1770,12 +1664,22 @@ var RealtimeCallHandler = class {
 				},
 				final: true
 			});
-		};
-		const submitFinalToolResult = async (result) => {
-			if (!delegation) await bridge.submitToolResult(bridgeCallId, result);
-			emitFinalToolEvent(result);
 			return result;
 		};
+		const invokeHandler = async (handlerArgs, context) => {
+			console.log(`[voice-call] realtime tool call executing callId=${callId} tool=${name} hasHandler=${Boolean(handler)}`);
+			try {
+				return handler ? await handler(handlerArgs, callId, context) : { error: `Tool "${name}" not available` };
+			} catch (error) {
+				return buildRealtimeVoiceAgentErrorProviderResult(error);
+			}
+		};
+		const logResult = (result) => {
+			const failed = hasResultError(result);
+			const error = failed ? formatErrorMessage(result.error ?? "unknown") : void 0;
+			console.log(`[voice-call] realtime tool call completed callId=${callId} tool=${name} status=${failed ? "error" : "ok"} elapsedMs=${Date.now() - startedAt}${error ? ` error=${error}` : ""}`);
+			return failed;
+		};
 		const submitWorkingResponse = async () => {
 			if (!delegation && handler && name === REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME && bridge.bridge.supportsToolResultContinuation && !this.config.fastContext.enabled) {
 				await bridge.submitToolResult(bridgeCallId, buildRealtimeVoiceAgentConsultWorkingResponse("caller"), { willContinue: true });
@@ -1826,17 +1730,12 @@ var RealtimeCallHandler = class {
 				return await submitFinalToolResult(outcome.result);
 			}
 			const abortController = new AbortController();
-			let releaseCancellation = () => {};
-			const cancellation = new Promise((resolve) => {
-				releaseCancellation = resolve;
-			});
-			let completeConsult = (_result) => {};
+			const { promise: cancellation, resolve: releaseCancellation } = createDeferred();
+			const { promise: consult, resolve: completeConsult } = createDeferred();
 			const state = {
 				owner: bridge,
 				startedAt,
-				promise: new Promise((resolve) => {
-					completeConsult = resolve;
-				}),
+				promise: consult,
 				cancellation,
 				get cancelled() {
 					return abortController.signal.aborted;
@@ -1861,8 +1760,7 @@ var RealtimeCallHandler = class {
 					};
 					state.partialUserTranscript = context.partialUserTranscript;
 					const handlerArgs = withFallbackConsultQuestion(args, context.partialUserTranscript);
-					console.log(`[voice-call] realtime tool call executing callId=${callId} tool=${name} hasHandler=${Boolean(handler)}`);
-					return !handler ? { error: `Tool "${name}" not available` } : await handler(handlerArgs, callId, context);
+					return await invokeHandler(handlerArgs, context);
 				} catch (error) {
 					return buildRealtimeVoiceAgentErrorProviderResult(error);
 				}
@@ -1871,9 +1769,7 @@ var RealtimeCallHandler = class {
 				const outcome = await waitForNativeConsult(state);
 				if (outcome.kind === "cancelled") return;
 				const result = outcome.result;
-				const failed = hasResultError(result);
-				const error = failed ? formatErrorMessage(result.error ?? "unknown") : void 0;
-				console.log(`[voice-call] realtime tool call completed callId=${callId} tool=${name} status=${failed ? "error" : "ok"} elapsedMs=${Date.now() - startedAt}${error ? ` error=${error}` : ""}`);
+				const failed = logResult(result);
 				await submitFinalToolResult(result);
 				if (!failed) this.consumePartialUserTranscript(callId, userTranscriptOwner, state.partialUserTranscript);
 				return result;
@@ -1882,16 +1778,8 @@ var RealtimeCallHandler = class {
 				if (this.nativeConsultsInFlightByCallId.get(callId) === state) this.nativeConsultsInFlightByCallId.delete(callId);
 			}
 		}
-		console.log(`[voice-call] realtime tool call executing callId=${callId} tool=${name} hasHandler=${Boolean(handler)}`);
-		const context = { partialUserTranscript: this.resolveUserTranscriptContext(callId, userTranscriptOwner) };
-		let result;
-		try {
-			result = !handler ? { error: `Tool "${name}" not available` } : await handler(args, callId, context);
-		} catch (error) {
-			result = buildRealtimeVoiceAgentErrorProviderResult(error);
-		}
-		const error = hasResultError(result) ? formatErrorMessage(result.error ?? "unknown") : void 0;
-		console.log(`[voice-call] realtime tool call completed callId=${callId} tool=${name} status=${error === void 0 ? "ok" : "error"} elapsedMs=${Date.now() - startedAt}${error ? ` error=${error}` : ""}`);
+		const result = await invokeHandler(args, { partialUserTranscript: this.resolveUserTranscriptContext(callId, userTranscriptOwner) });
+		logResult(result);
 		return await submitFinalToolResult(result);
 	}
 };

package/dist/.setup/{response-generator-BzKrJMEb.mjs → response-generator-DAhbXEta.mjs}

@@ -1,6 +1,6 @@
-import { c as resolveVoiceCallSessionKey } from "./config-Ds7xoZbI.mjs";
-import { _ as resolveCallAgentId, g as resolveVoiceResponseModel } from "./runtime-entry-BNOQRXii.mjs";
-import { filterStringEntries, isRecord, normalizeLowercaseStringOrEmpty, normalizeStringEntries } from "openclaw/plugin-sdk/string-coerce-runtime";
+import { c as resolveVoiceCallSessionKey } from "./config-CX9J-053.mjs";
+import { m as resolveCallAgentId, p as resolveVoiceResponseModel } from "./runtime-entry-CbSKxSKk.mjs";
+import { normalizeLowercaseStringOrEmpty, normalizeStringEntries } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { resolveAgentConfig } from "openclaw/plugin-sdk/agent-scope-runtime";
 import { isValidAgentHarnessSessionStoreEntry } from "openclaw/plugin-sdk/session-store-runtime";
 import crypto from "node:crypto";
@@ -12,15 +12,6 @@ import { ModelSelectionLockedError, applyModelOverrideWithAuthProfileCompatibili
 * Voice call response generator - uses the embedded OpenClaw agent for tool support.
 * Routes voice responses through the same agent infrastructure as messaging.
 */
-function readExplicitToolsAllow(value) {
-	if (!isRecord(value)) return;
-	const allow = value.allow;
-	if (!Array.isArray(allow)) return;
-	return filterStringEntries(allow);
-}
-function resolveVoiceAgentToolsAllow(config, agentId) {
-	return readExplicitToolsAllow(resolveAgentConfig(config, agentId)?.tools);
-}
 const VOICE_SPOKEN_OUTPUT_CONTRACT = [
 	"Output format requirements:",
 	"- Return only valid JSON in this exact shape: {\"spoken\":\"...\"}",
@@ -93,12 +84,7 @@ function tryParseSpokenJson(text) {
 }
 function isLikelyMetaReasoningParagraph(paragraph) {
 	const lower = normalizeLowercaseStringOrEmpty(paragraph);
-	if (!lower) return false;
-	if (lower.startsWith("thinking process")) return true;
-	if (lower.startsWith("reasoning:") || lower.startsWith("analysis:")) return true;
-	if (lower.startsWith("the user ") && (lower.includes("i should") || lower.includes("i need to") || lower.includes("i will"))) return true;
-	if (lower.includes("this is a natural continuation of the conversation") || lower.includes("keep the conversation flowing")) return true;
-	return false;
+	return lower.startsWith("thinking process") || lower.startsWith("reasoning:") || lower.startsWith("analysis:") || lower.startsWith("the user ") && (lower.includes("i should") || lower.includes("i need to") || lower.includes("i will")) || lower.includes("this is a natural continuation of the conversation") || lower.includes("keep the conversation flowing");
 }
 function sanitizePlainSpokenText(text) {
 	const withoutCodeFences = text.replace(/```[\s\S]*?```/g, " ").trim();
@@ -146,13 +132,8 @@ function resolveVoiceSandboxSessionKey(agentId, sessionKey) {
 */
 async function generateVoiceResponse(params) {
 	const { voiceConfig, callId, sessionKey, from, senderIsOwner, transcript, userMessage, coreConfig, agentRuntime, onEarlyText } = params;
-	if (!coreConfig) return {
-		text: null,
-		deliveredEarly: false,
-		error: "Core config unavailable for voice response"
-	};
 	const cfg = coreConfig;
-	const agentId = resolveCallAgentId({ agentId: params.agentId }, voiceConfig);
+	const agentId = resolveCallAgentId(params);
 	const resolvedSessionKey = resolveVoiceCallSessionKey({
 		config: {
 			...voiceConfig,
@@ -163,7 +144,7 @@ async function generateVoiceResponse(params) {
 		explicitSessionKey: sessionKey,
 		coreSession: coreConfig.session
 	});
-	const toolsAllow = resolveVoiceAgentToolsAllow(cfg, agentId);
+	const toolsAllow = resolveAgentConfig(cfg, agentId)?.tools?.allow;
 	const storePath = agentRuntime.session.resolveStorePath(cfg.session?.store, { agentId });
 	try {
 		return await agentRuntime.session.runWithWorkAdmission({