@openclaw/discord 2026.7.2-beta.2 → 2026.7.2-beta.4

package/dist/{manager.runtime-w6VVYQMl.js → manager.runtime-DVvlH8NR.js}

@@ -1,8 +1,8 @@
-import { A as isUnknownDiscordVoiceStateError, Bt as getGuildVoiceState, S as ResumedListener, d as parseDiscordTarget, p as discord_exports, w as VoiceStateUpdateListener, x as ReadyListener } from "./chunk-CHpY_lsO.js";
-import { a as normalizeDiscordSlug, h as resolveDiscordOwnerAccess, x as formatDiscordUserTag } from "./allow-list-B7qOH5xu.js";
+import { O as isUnknownDiscordVoiceStateError, Rt as getGuildVoiceState, S as VoiceStateUpdateListener, b as ResumedListener, d as discord_exports, l as parseDiscordTarget, y as ReadyListener } from "./chunk-BrZocpfs.js";
 import { t as getDiscordRuntime } from "./runtime-DgnVQ7zW.js";
-import { a as formatMention } from "./send.outbound-Cvulb-k_.js";
-import { c as resolveDiscordVoiceAccess, h as buildDiscordGroupSystemPrompt, l as authorizeDiscordVoiceIngress, p as resolveDiscordVoiceEnabled } from "./provider-PRlF_lZm.js";
+import { a as normalizeDiscordSlug, h as resolveDiscordOwnerAccess, x as formatDiscordUserTag } from "./allow-list-DzKrtfah.js";
+import { a as formatMention } from "./send.outbound-BbC2qjpH.js";
+import { g as buildDiscordGroupSystemPrompt, l as resolveDiscordVoiceAccess, m as resolveDiscordVoiceEnabled, u as authorizeDiscordVoiceIngress } from "./provider-CfiChtjf.js";
 import { createRequire } from "node:module";
 import { resolveAgentRoute } from "openclaw/plugin-sdk/routing";
 import { asBoolean, normalizeOptionalString, uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";
@@ -16,15 +16,14 @@ import { stripInlineDirectiveTagsForDisplay } from "openclaw/plugin-sdk/text-chu
 import { resolveFfmpegBin } from "openclaw/plugin-sdk/media-runtime";
 import path from "node:path";
 import { resolvePreferredOpenClawTmpDir, tempWorkspace } from "openclaw/plugin-sdk/temp-path";
-import { REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME, REALTIME_VOICE_AGENT_CONTROL_TOOL, REALTIME_VOICE_AGENT_CONTROL_TOOL_NAME, REALTIME_VOICE_AUDIO_FORMAT_PCM16_24KHZ, buildRealtimeVoiceAgentConsultChatMessage, buildRealtimeVoiceAgentConsultPolicyInstructions, classifySkippableRealtimeVoiceConsultTranscript, controlRealtimeVoiceAgentRun, createRealtimeVoiceAgentTalkbackQueue, createRealtimeVoiceBridgeSession, createRealtimeVoiceForcedConsultCoordinator, createRealtimeVoiceOutputActivityTracker, createRealtimeVoiceTurnContextTracker, matchRealtimeVoiceActivationName, matchRealtimeVoiceConsultQuestions, normalizeSupportedRealtimeVoiceActivationName, parseRealtimeVoiceAgentControlToolArgs, resamplePcm, resolveConfiguredRealtimeVoiceProvider, resolveRealtimeVoiceAgentConsultToolPolicy, resolveRealtimeVoiceAgentConsultTools, resolveRealtimeVoiceAgentConsultToolsAllow, shouldAutoControlRealtimeVoiceAgentText, sortRealtimeVoiceActivationNames } from "openclaw/plugin-sdk/realtime-voice";
-import { agentCommandFromIngress, getTtsProvider, resolveAgentDir, resolveTtsConfig, resolveTtsPrefsPath } from "openclaw/plugin-sdk/agent-runtime";
+import { REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME, REALTIME_VOICE_AGENT_CONTROL_TOOL, REALTIME_VOICE_AGENT_CONTROL_TOOL_NAME, REALTIME_VOICE_AUDIO_FORMAT_PCM16_24KHZ, buildRealtimeVoiceAgentConsultChatMessage, buildRealtimeVoiceAgentConsultPolicyInstructions, classifySkippableRealtimeVoiceConsultTranscript, controlRealtimeVoiceAgentRun, createRealtimeVoiceSessionHarness, createRealtimeVoiceTurnContextTracker, matchRealtimeVoiceActivationName, matchRealtimeVoiceConsultQuestions, normalizeSupportedRealtimeVoiceActivationName, parseRealtimeVoiceAgentControlToolArgs, resamplePcm, resolveConfiguredRealtimeVoiceProvider, resolveRealtimeVoiceAgentConsultToolPolicy, resolveRealtimeVoiceAgentConsultTools, resolveRealtimeVoiceAgentConsultToolsAllow, shouldAutoControlRealtimeVoiceAgentText, sortRealtimeVoiceActivationNames } from "openclaw/plugin-sdk/realtime-voice";
+import { agentCommandFromIngress, resolveAgentDir } from "openclaw/plugin-sdk/agent-runtime";
 import { enqueueSystemEvent } from "openclaw/plugin-sdk/system-event-runtime";
 import { spawn } from "node:child_process";
 import { PassThrough, Readable, Transform, pipeline } from "node:stream";
 import { StringDecoder } from "node:string_decoder";
 import { Application, OpusError, createDecoder, createEncoder } from "libopus-wasm";
 import { resolveRealtimeBootstrapContextInstructions } from "openclaw/plugin-sdk/realtime-bootstrap-context";
-import { parseTtsDirectives } from "openclaw/plugin-sdk/speech";
 //#region extensions/discord/src/voice/audio.ts
 const SAMPLE_RATE = 48e3;
 const CHANNELS = 2;
@@ -485,7 +484,7 @@ const MAX_ADDITIONAL_PARTICIPANTS = 256;
 function normalizeLabel(value) {
 	if (typeof value !== "string") return;
 	const normalized = value.replace(/\s+/g, " ").trim();
-	return normalized ? normalized.slice(0, 100) : void 0;
+	return normalized ? truncateUtf16Safe(normalized, 100) : void 0;
 }
 function memberLabel(state) {
 	return normalizeLabel(state.member?.nick) ?? normalizeLabel(state.member?.user?.global_name) ?? normalizeLabel(state.member?.user?.username);
@@ -926,6 +925,7 @@ const DISCORD_REALTIME_WAKE_ACKS = [
 	"Got it.",
 	"One sec."
 ];
+const discordRealtimeTalkPayload = () => ({});
 const REALTIME_PCM16_BYTES_PER_SAMPLE = 2;
 const DISCORD_RAW_PCM_FRAME_BYTES = 3840;
 const DISCORD_REALTIME_OUTPUT_PREROLL_FRAMES = 25;
@@ -1041,9 +1041,6 @@ function extractDiscordExactSpeechConsultText(args) {
 function normalizeControlSpeechText(text) {
 	return text.toLowerCase().replace(/\s+/g, " ").trim();
 }
-function matchesPendingAgentProxyQuestion(consultMessage, question) {
-	return matchRealtimeVoiceConsultQuestions(consultMessage, question);
-}
 var DiscordRealtimeVoiceSession = class {
 	constructor(params) {
 		this.params = params;
@@ -1054,17 +1051,11 @@ var DiscordRealtimeVoiceSession = class {
 		this.consultPolicy = "auto";
 		this.wakeNamePolicy = "never";
 		this.wakeNames = [];
-		this.forcedConsults = createRealtimeVoiceForcedConsultCoordinator({
-			limit: DISCORD_REALTIME_RECENT_AGENT_PROXY_CONSULT_LIMIT,
-			nativeDedupeMs: DISCORD_REALTIME_RECENT_AGENT_PROXY_CONSULT_TTL_MS,
-			questionsMatch: matchesPendingAgentProxyQuestion
-		});
 		this.speakerTurns = createRealtimeVoiceTurnContextTracker({
 			limit: DISCORD_REALTIME_PENDING_SPEAKER_CONTEXT_LIMIT,
 			ignoredContextTtlMs: DISCORD_REALTIME_IGNORED_WAKE_NAME_CONTEXT_TTL_MS,
 			deferUntilAudio: true
 		});
-		this.outputActivity = createRealtimeVoiceOutputActivityTracker();
 		this.outputPacedBuffer = Buffer.alloc(0);
 		this.queuedExactSpeechMessages = [];
 		this.exactSpeechResponseActive = false;
@@ -1077,21 +1068,41 @@ var DiscordRealtimeVoiceSession = class {
 			this.resetOutputStream("player-idle");
 			if (hadOutputAudio) this.completeExactSpeechResponse("player-idle");
 		};
-		this.talkback = createRealtimeVoiceAgentTalkbackQueue({
-			debounceMs: this.realtimeConfig?.debounceMs ?? DISCORD_REALTIME_TALKBACK_DEBOUNCE_MS,
-			isStopped: () => this.stopped,
-			logger: logger$3,
-			logPrefix: "[discord] realtime agent",
-			responseStyle: "Brief, natural spoken answer for a Discord voice channel.",
-			fallbackText: DISCORD_REALTIME_FALLBACK_TEXT,
-			consult: async ({ question, responseStyle, metadata }) => {
-				const context = isDiscordRealtimeSpeakerContext(metadata) ? metadata : void 0;
-				return { text: await this.runAgentTurn({
-					context,
-					message: formatVoiceIngressPrompt([question, responseStyle ? `Spoken style: ${responseStyle}` : void 0].filter(Boolean).join("\n\n"), context?.speakerLabel ?? "Discord voice speaker")
-				}) };
+		this.harness = createRealtimeVoiceSessionHarness({
+			talk: {
+				sessionId: `discord:${this.params.entry.voiceSessionKey}:realtime`,
+				mode: "realtime",
+				transport: "gateway-relay",
+				brain: "agent-consult"
+			},
+			talkPayloads: {
+				turnStarted: discordRealtimeTalkPayload,
+				turnEnded: discordRealtimeTalkPayload,
+				inputAudioDelta: discordRealtimeTalkPayload,
+				outputAudioStarted: discordRealtimeTalkPayload,
+				outputAudioDelta: discordRealtimeTalkPayload,
+				outputAudioDone: discordRealtimeTalkPayload
+			},
+			talkback: {
+				debounceMs: this.realtimeConfig?.debounceMs ?? DISCORD_REALTIME_TALKBACK_DEBOUNCE_MS,
+				logger: logger$3,
+				logPrefix: "[discord] realtime agent",
+				responseStyle: "Brief, natural spoken answer for a Discord voice channel.",
+				fallbackText: DISCORD_REALTIME_FALLBACK_TEXT,
+				consult: async ({ question, responseStyle, metadata }) => {
+					const context = isDiscordRealtimeSpeakerContext(metadata) ? metadata : void 0;
+					return { text: await this.runAgentTurn({
+						context,
+						message: formatVoiceIngressPrompt([question, responseStyle ? `Spoken style: ${responseStyle}` : void 0].filter(Boolean).join("\n\n"), context?.speakerLabel ?? "Discord voice speaker")
+					}) };
+				},
+				deliver: (text) => this.enqueueExactSpeechMessage(text)
 			},
-			deliver: (text) => this.enqueueExactSpeechMessage(text)
+			forcedConsults: {
+				limit: DISCORD_REALTIME_RECENT_AGENT_PROXY_CONSULT_LIMIT,
+				nativeDedupeMs: DISCORD_REALTIME_RECENT_AGENT_PROXY_CONSULT_TTL_MS,
+				questionsMatch: matchRealtimeVoiceConsultQuestions
+			}
 		});
 	}
 	async connect() {
@@ -1134,7 +1145,7 @@ var DiscordRealtimeVoiceSession = class {
 			toolPolicy,
 			consultPolicy
 		});
-		this.bridge = createRealtimeVoiceBridgeSession({
+		this.bridge = this.harness.createBridge({
 			provider: resolved.provider,
 			cfg: this.params.cfg,
 			providerConfig: resolved.providerConfig,
@@ -1147,7 +1158,7 @@ var DiscordRealtimeVoiceSession = class {
 			audioSink: {
 				isOpen: () => !this.stopped,
 				sendAudio: (audio) => this.sendOutputAudio(audio),
-				clearAudio: () => this.clearOutputAudio("provider-clear-audio")
+				clearAudio: () => this.harness.flushOutput(() => this.clearOutputAudio("provider-clear-audio"))
 			},
 			onTranscript: (role, text, isFinal) => {
 				if (isFinal && text.trim()) logger$3.info(`discord voice: realtime ${role} transcript (${text.length} chars): ${formatVoiceLogPreview(text)}`);
@@ -1194,8 +1205,7 @@ var DiscordRealtimeVoiceSession = class {
 	close() {
 		this.stopped = true;
 		this.flushSuppressedRealtimeErrors();
-		this.talkback.close();
-		this.forcedConsults.clear();
+		this.harness.close();
 		this.speakerTurns.clear();
 		this.queuedExactSpeechMessages = [];
 		this.exactSpeechResponseActive = false;
@@ -1278,11 +1288,11 @@ var DiscordRealtimeVoiceSession = class {
 			return;
 		}
 		if (!this.hasInterruptibleOutputAudio()) {
-			logger$3.info(`discord voice: realtime barge-in ignored reason=${reason} outputActive=false guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} playbackChunks=${this.outputAudioChunks()}`);
+			logger$3.info(`discord voice: realtime barge-in ignored reason=${reason} outputActive=false guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} playbackChunks=${this.harness.outputActivity.snapshot().chunks}`);
 			return;
 		}
-		logger$3.info(`discord voice: realtime barge-in requested reason=${reason} guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} outputAudioMs=${this.outputAudioMs()} outputActive=${this.isOutputAudioActive()} playbackChunks=${this.outputAudioChunks()}`);
-		this.bridge?.handleBargeIn({ audioPlaybackActive: true });
+		logger$3.info(`discord voice: realtime barge-in requested reason=${reason} guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} outputAudioMs=${this.outputAudioMs()} outputActive=${this.isOutputAudioActive()} playbackChunks=${this.harness.outputActivity.snapshot().chunks}`);
+		this.harness.handleBargeIn({ audioPlaybackActive: true }, () => {});
 	}
 	isBargeInEnabled() {
 		if (this.isWakeNameRequired()) return false;
@@ -1293,8 +1303,9 @@ var DiscordRealtimeVoiceSession = class {
 		});
 	}
 	hasInterruptibleOutputAudio() {
-		this.syncOutputAudioTimestamp();
-		return this.outputActivity.isInterruptible(this.isOutputStreamActive());
+		this.bridge?.setMediaTimestamp(this.outputAudioMs());
+		const streamActive = Boolean(this.outputStream && !this.outputStream.destroyed);
+		return this.harness.outputActivity.isInterruptible(streamActive);
 	}
 	get realtimeConfig() {
 		return this.params.discordConfig.voice?.realtime;
@@ -1308,14 +1319,14 @@ var DiscordRealtimeVoiceSession = class {
 	sendOutputAudio(realtimePcm24kMono) {
 		const discordPcm = convertRealtimePcm24kMonoToDiscordPcm48kStereo(realtimePcm24kMono);
 		if (discordPcm.length === 0) return;
-		this.syncOutputAudioTimestamp();
-		if (this.outputActivity.snapshot().streamEnding) {
+		this.bridge?.setMediaTimestamp(this.outputAudioMs());
+		if (this.harness.outputActivity.snapshot().streamEnding) {
 			logVoiceVerbose(`realtime output audio ignored after stream ending: guild ${this.params.entry.guildId} channel ${this.params.entry.channelId}`);
 			return;
 		}
 		const stream = this.ensureOutputStream();
 		if (this.exactSpeechResponseActive) this.exactSpeechAudioStarted = true;
-		this.outputActivity.markAudio({
+		this.harness.outputActivity.markAudio({
 			audioMs: pcm16MonoDurationMs(realtimePcm24kMono, REALTIME_VOICE_AUDIO_FORMAT_PCM16_24KHZ.sampleRateHz),
 			sourceAudioBytes: realtimePcm24kMono.length,
 			sinkAudioBytes: discordPcm.length
@@ -1327,9 +1338,9 @@ var DiscordRealtimeVoiceSession = class {
 		const stream = new PassThrough({ highWaterMark: DISCORD_RAW_PCM_FRAME_BYTES * 128 });
 		this.outputStream = stream;
 		this.outputPacedBuffer = Buffer.alloc(0);
-		this.outputActivity.markStreamOpened();
+		this.harness.outputActivity.markStreamOpened();
 		stream.once("close", () => {
-			if (this.outputActivity.snapshot().playbackStarted) return;
+			if (this.harness.outputActivity.snapshot().playbackStarted) return;
 			this.handleOutputStreamClosed(stream, "stream-close");
 		});
 		return stream;
@@ -1339,11 +1350,12 @@ var DiscordRealtimeVoiceSession = class {
 		this.logOutputAudioStopped(reason);
 		this.clearOutputPlaybackWatchdog();
 		this.outputStream = null;
-		this.resetOutputAudioStats();
+		this.outputPacedBuffer = Buffer.alloc(0);
+		this.harness.outputActivity.reset();
 		this.completeExactSpeechResponse(reason);
 	}
 	queueOutputAudio(stream, discordPcm) {
-		if (this.outputActivity.snapshot().playbackStarted) {
+		if (this.harness.outputActivity.snapshot().playbackStarted) {
 			stream.write(discordPcm);
 			return;
 		}
@@ -1351,7 +1363,7 @@ var DiscordRealtimeVoiceSession = class {
 		if (this.outputPacedBuffer.length >= DISCORD_RAW_PCM_FRAME_BYTES * DISCORD_REALTIME_OUTPUT_PREROLL_FRAMES) this.startOutputPlayback(stream);
 	}
 	startOutputPlayback(stream) {
-		if (this.outputActivity.snapshot().playbackStarted || stream.destroyed) return;
+		if (this.harness.outputActivity.snapshot().playbackStarted || stream.destroyed) return;
 		const voiceSdk = loadDiscordVoiceSdk();
 		const opusStream = createDiscordOpusEncodeStream();
 		opusStream.on("error", (err) => {
@@ -1370,9 +1382,9 @@ var DiscordRealtimeVoiceSession = class {
 		}
 		const resource = voiceSdk.createAudioResource(opusStream, { inputType: voiceSdk.StreamType.Opus });
 		this.params.entry.player.play(resource);
-		this.outputActivity.markPlaybackStarted();
+		this.harness.outputActivity.markPlaybackStarted();
 		const realtimeConfig = this.realtimeConfig;
-		logger$3.info(`discord voice: realtime audio playback started guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} mode=${this.params.mode} model=${realtimeConfig?.model ?? "provider-default"} voice=${realtimeConfig?.voice ?? "provider-default"}`);
+		logger$3.info(`discord voice: realtime audio playback started guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} mode=${this.params.mode} model=${realtimeConfig?.model ?? "provider-default"} voice=${realtimeConfig?.speakerVoice ?? realtimeConfig?.speakerVoiceId ?? "provider-default"}`);
 	}
 	clearOutputAudio(reason = "clear") {
 		this.resetOutputStream(reason);
@@ -1384,15 +1396,15 @@ var DiscordRealtimeVoiceSession = class {
 		this.logOutputAudioStopped(reason);
 		this.outputStream = null;
 		this.outputPacedBuffer = Buffer.alloc(0);
-		this.resetOutputAudioStats();
+		this.harness.outputActivity.reset();
 		stream?.end();
 		stream?.destroy();
 	}
 	finishOutputAudioStream(reason, { playBuffered = true } = {}) {
 		const stream = this.outputStream;
-		if (!stream || stream.destroyed || this.outputActivity.snapshot().streamEnding) return;
-		this.outputActivity.markStreamEnding();
-		logger$3.info(`discord voice: realtime audio playback finishing reason=${reason} guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} audioMs=${this.outputAudioMs()} chunks=${this.outputAudioChunks()}`);
+		if (!stream || stream.destroyed || this.harness.outputActivity.snapshot().streamEnding) return;
+		this.harness.outputActivity.markStreamEnding();
+		logger$3.info(`discord voice: realtime audio playback finishing reason=${reason} guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} audioMs=${this.outputAudioMs()} chunks=${this.harness.outputActivity.snapshot().chunks}`);
 		if (playBuffered) {
 			this.startOutputPlayback(stream);
 			this.scheduleOutputPlaybackWatchdog(reason, stream);
@@ -1406,7 +1418,7 @@ var DiscordRealtimeVoiceSession = class {
 	}
 	scheduleOutputPlaybackWatchdog(reason, stream) {
 		this.clearOutputPlaybackWatchdog();
-		const timeoutMs = this.outputActivity.playbackWatchdogDelayMs({ marginMs: DISCORD_REALTIME_OUTPUT_PLAYBACK_WATCHDOG_MARGIN_MS });
+		const timeoutMs = this.harness.outputActivity.playbackWatchdogDelayMs({ marginMs: DISCORD_REALTIME_OUTPUT_PLAYBACK_WATCHDOG_MARGIN_MS });
 		if (timeoutMs === void 0) return;
 		this.outputPlaybackWatchdog = setTimeout(() => {
 			this.outputPlaybackWatchdog = void 0;
@@ -1415,7 +1427,7 @@ var DiscordRealtimeVoiceSession = class {
 				this.completeExactSpeechResponse("playback-watchdog");
 				return;
 			}
-			logger$3.warn(`discord voice: realtime audio playback watchdog fired reason=${reason} guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} audioMs=${this.outputAudioMs()} elapsedMs=${this.outputActivity.elapsedPlaybackMs()}`);
+			logger$3.warn(`discord voice: realtime audio playback watchdog fired reason=${reason} guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} audioMs=${this.outputAudioMs()} elapsedMs=${this.harness.outputActivity.elapsedPlaybackMs()}`);
 			this.clearOutputAudio("playback-watchdog");
 			this.completeExactSpeechResponse("playback-watchdog");
 		}, timeoutMs);
@@ -1456,11 +1468,10 @@ var DiscordRealtimeVoiceSession = class {
 		if (this.stopped || !trimmed) return;
 		this.queuedExactSpeechMessages = [];
 		this.completeExactSpeechResponse("active-run-control", { drain: false });
-		this.bridge?.handleBargeIn?.({
+		this.harness.handleBargeIn({
 			audioPlaybackActive: true,
 			force: true
-		});
-		this.clearOutputAudio("active-run-control");
+		}, () => this.clearOutputAudio("active-run-control"));
 		this.lastControlSpeech = {
 			normalizedText: normalizeControlSpeechText(trimmed),
 			sentAt: Date.now(),
@@ -1479,11 +1490,10 @@ var DiscordRealtimeVoiceSession = class {
 		recent.assistantTranscriptCount += 1;
 		if (recent.assistantTranscriptCount <= 1) return;
 		logger$3.info(`discord voice: realtime duplicate active-run control speech suppressed guild=${this.params.entry.guildId} channel=${this.params.entry.channelId}`);
-		this.bridge?.handleBargeIn?.({
+		this.harness.handleBargeIn({
 			audioPlaybackActive: true,
 			force: true
-		});
-		this.clearOutputAudio("duplicate-active-run-control");
+		}, () => this.clearOutputAudio("duplicate-active-run-control"));
 	}
 	completeExactSpeechResponse(reason, options) {
 		if (!this.exactSpeechResponseActive && this.queuedExactSpeechMessages.length === 0) return;
@@ -1500,32 +1510,19 @@ var DiscordRealtimeVoiceSession = class {
 		this.sendExactSpeechMessage(next);
 	}
 	logOutputAudioStopped(reason) {
-		const activity = this.outputActivity.snapshot();
+		const activity = this.harness.outputActivity.snapshot();
 		const audioMs = Math.floor(activity.audioMs);
 		const chunks = activity.chunks;
 		const discordBytes = activity.sinkAudioBytes;
 		const realtimeBytes = activity.sourceAudioBytes;
-		const elapsedMs = this.outputActivity.elapsedPlaybackMs();
+		const elapsedMs = this.harness.outputActivity.elapsedPlaybackMs();
 		if (this.outputStream || chunks > 0 || audioMs > 0) logger$3.info(`discord voice: realtime audio playback stopped reason=${reason} guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} audioMs=${audioMs} elapsedMs=${elapsedMs} chunks=${chunks} discordBytes=${discordBytes} realtimeBytes=${realtimeBytes}`);
 	}
-	resetOutputAudioStats() {
-		this.outputPacedBuffer = Buffer.alloc(0);
-		this.outputActivity.reset();
-	}
-	syncOutputAudioTimestamp() {
-		this.bridge?.setMediaTimestamp(this.outputAudioMs());
-	}
 	outputAudioMs() {
-		return Math.floor(this.outputActivity.snapshot().audioMs);
-	}
-	outputAudioChunks() {
-		return this.outputActivity.snapshot().chunks;
-	}
-	isOutputStreamActive() {
-		return Boolean(this.outputStream && !this.outputStream.destroyed);
+		return Math.floor(this.harness.outputActivity.snapshot().audioMs);
 	}
 	isOutputAudioActive() {
-		return this.outputActivity.isActive(this.isOutputStreamActive());
+		return this.harness.outputActivity.isActive(Boolean(this.outputStream && !this.outputStream.destroyed));
 	}
 	logSpeakerTurnClosed(turn) {
 		if (turn.closed || !turn.hasAudio) return;
@@ -1573,8 +1570,8 @@ var DiscordRealtimeVoiceSession = class {
 			return;
 		}
 		logger$3.info(`discord voice: realtime consult requested call=${callId || "unknown"} voiceSession=${this.params.entry.voiceSessionKey} supervisorSession=${this.params.entry.route.sessionKey} agent=${this.params.entry.route.agentId} question=${formatVoiceLogPreview(consultMessage)}`);
-		const nativeConsult = this.forcedConsults.recordNativeConsult(event.args, callId);
-		if (nativeConsult.kind === "already_delivered" && this.forcedConsults.isCancelled(nativeConsult.handle)) {
+		const nativeConsult = this.harness.forcedConsults.recordNativeConsult(event.args, callId);
+		if (nativeConsult.kind === "already_delivered" && this.harness.forcedConsults.isCancelled(nativeConsult.handle)) {
 			await this.submitTerminalRealtimeToolResult(callId, session, {
 				status: "cancelled",
 				message: "OpenClaw cancelled this consult before completion. Do not restart it."
@@ -1582,7 +1579,7 @@ var DiscordRealtimeVoiceSession = class {
 			return;
 		}
 		const pendingConsult = nativeConsult.kind === "pending" ? nativeConsult.handle : void 0;
-		if (pendingConsult) this.forcedConsults.rememberQuestion(pendingConsult, consultMessage);
+		if (pendingConsult) this.harness.forcedConsults.rememberQuestion(pendingConsult, consultMessage);
 		let context = pendingConsult?.context?.speaker;
 		let recent = pendingConsult;
 		if (!context) {
@@ -1690,7 +1687,7 @@ var DiscordRealtimeVoiceSession = class {
 			logger$3.warn(`discord voice: realtime active-run control failed; falling back to normal transcript handling: ${formatErrorMessage(error)}`);
 		});
 		if (control?.handled) {
-			if (pendingForcedConsult) this.forcedConsults.remove(pendingForcedConsult);
+			if (pendingForcedConsult) this.harness.forcedConsults.remove(pendingForcedConsult);
 			this.logAgentControlResult(control.result);
 			if (control.speakText) this.speakControlResult(control.speakText);
 			return;
@@ -1700,7 +1697,7 @@ var DiscordRealtimeVoiceSession = class {
 			if (pendingForcedConsult) this.schedulePreparedForcedAgentProxyConsult(pendingForcedConsult);
 			return;
 		}
-		this.talkback.enqueue(acceptedText, forcedSpeakerContext ?? this.consumePendingSpeakerContext());
+		this.harness.talkback?.enqueue(acceptedText, forcedSpeakerContext ?? this.consumePendingSpeakerContext());
 	}
 	handlePartialUserTranscript(text) {
 		if (!this.isWakeNameRequired() || this.wakeNameAckedForTurn) return;
@@ -1787,28 +1784,28 @@ var DiscordRealtimeVoiceSession = class {
 			logger$3.warn("discord voice: realtime forced agent consult has no speaker context");
 			return;
 		}
-		return this.forcedConsults.prepare(question, { context: { speaker: context } });
+		return this.harness.forcedConsults.prepare(question, { context: { speaker: context } });
 	}
 	schedulePreparedForcedAgentProxyConsult(pending) {
-		this.forcedConsults.schedule(pending, DISCORD_REALTIME_FORCED_CONSULT_FALLBACK_DELAY_MS, (handle) => void this.runForcedAgentProxyConsult(handle));
+		this.harness.forcedConsults.schedule(pending, DISCORD_REALTIME_FORCED_CONSULT_FALLBACK_DELAY_MS, (handle) => void this.runForcedAgentProxyConsult(handle));
 	}
 	async runForcedAgentProxyConsult(pending) {
-		this.forcedConsults.markStarted(pending);
+		this.harness.forcedConsults.markStarted(pending);
 		const state = pending.context;
 		if (!state) {
-			this.forcedConsults.markCancelled(pending);
+			this.harness.forcedConsults.markCancelled(pending);
 			return;
 		}
 		const context = state.speaker;
 		const { question } = pending;
 		if (this.stopped) {
-			this.forcedConsults.markCancelled(pending);
+			this.harness.forcedConsults.markCancelled(pending);
 			return;
 		}
 		const startedAt = Date.now();
 		logger$3.info(`discord voice: realtime forced agent consult starting chars=${question.length} voiceSession=${this.params.entry.voiceSessionKey} supervisorSession=${this.params.entry.route.sessionKey} agent=${this.params.entry.route.agentId} speaker=${context.speakerLabel} owner=${context.senderIsOwner}`);
 		logger$3.debug(`discord voice: realtime forced agent consult reason=${DISCORD_REALTIME_FORCED_CONSULT_REASON} consultPolicy=${this.consultPolicy} wakeNamePolicy=${this.wakeNamePolicy} requireWakeName=${this.isWakeNameRequired()} voiceSession=${this.params.entry.voiceSessionKey} supervisorSession=${this.params.entry.route.sessionKey} agent=${this.params.entry.route.agentId} speaker=${context.speakerLabel}`);
-		if (this.hasInterruptibleOutputAudio()) logger$3.info(`discord voice: realtime forced agent consult preserving active playback guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} outputAudioMs=${this.outputAudioMs()} outputActive=${this.isOutputAudioActive()} playbackChunks=${this.outputAudioChunks()}`);
+		if (this.hasInterruptibleOutputAudio()) logger$3.info(`discord voice: realtime forced agent consult preserving active playback guild=${this.params.entry.guildId} channel=${this.params.entry.channelId} outputAudioMs=${this.outputAudioMs()} outputActive=${this.isOutputAudioActive()} playbackChunks=${this.harness.outputActivity.snapshot().chunks}`);
 		state.handledByForcedPlayback = true;
 		try {
 			const promise = this.runAgentTurn({
@@ -1872,35 +1869,35 @@ var DiscordRealtimeVoiceSession = class {
 		return this.speakerTurns.hasAudioContext();
 	}
 	rememberRecentAgentProxyConsultContext(question, context, options = {}) {
-		const handle = this.forcedConsults.prepare(question, {
+		const handle = this.harness.forcedConsults.prepare(question, {
 			context: { speaker: context },
 			...options.id ? { id: options.id } : {}
 		});
 		if (!handle) throw new Error("Discord realtime consult context requires a non-empty question");
-		if (options.started) this.forcedConsults.markStarted(handle);
+		if (options.started) this.harness.forcedConsults.markStarted(handle);
 		return handle;
 	}
 	setRecentAgentProxyConsultPromise(recent, promise) {
 		const state = recent.context;
 		if (!state) return;
-		this.forcedConsults.markStarted(recent);
+		this.harness.forcedConsults.markStarted(recent);
 		state.promise = promise;
 		promise.then((text) => {
 			state.result = {
 				status: "fulfilled",
 				text
 			};
-			this.forcedConsults.markDelivered(recent);
+			this.harness.forcedConsults.markDelivered(recent);
 		}).catch((error) => {
 			state.result = {
 				status: "rejected",
 				error: formatErrorMessage(error)
 			};
-			this.forcedConsults.markDelivered(recent);
+			this.harness.forcedConsults.markDelivered(recent);
 		});
 	}
 	findRecentAgentProxyConsultContext(consultMessage) {
-		return this.forcedConsults.findRecent(consultMessage);
+		return this.harness.forcedConsults.findRecent(consultMessage);
 	}
 	async submitTerminalRealtimeToolResult(callId, session, result) {
 		if (session.bridge.supportsToolResultSuppression === false) {
@@ -1985,7 +1982,7 @@ function buildProviderConfigs(realtimeConfig) {
 function buildProviderConfigOverrides(realtimeConfig) {
 	const overrides = {
 		...realtimeConfig?.model ? { model: realtimeConfig.model } : {},
-		...realtimeConfig?.speakerVoice ? { voice: realtimeConfig.speakerVoice } : realtimeConfig?.speakerVoiceId ? { voice: realtimeConfig.speakerVoiceId } : realtimeConfig?.voice ? { voice: realtimeConfig.voice } : {},
+		...realtimeConfig?.speakerVoice ? { voice: realtimeConfig.speakerVoice } : realtimeConfig?.speakerVoiceId ? { voice: realtimeConfig.speakerVoiceId } : {},
 		...typeof realtimeConfig?.minBargeInAudioEndMs === "number" ? { minBargeInAudioEndMs: realtimeConfig.minBargeInAudioEndMs } : {}
 	};
 	return Object.keys(overrides).length > 0 ? overrides : void 0;
@@ -2118,47 +2115,6 @@ function sanitizeVoiceReplyTextForSpeech(text, speakerLabel) {
 }
 //#endregion
 //#region extensions/discord/src/voice/tts.ts
-function mergeTtsConfig(base, override) {
-	if (!override) return base;
-	const baseProviders = base.providers ?? {};
-	const overrideProviders = override.providers ?? {};
-	const mergedProviders = Object.fromEntries(uniqueStrings([...Object.keys(baseProviders), ...Object.keys(overrideProviders)]).map((providerId) => {
-		const baseProvider = baseProviders[providerId] ?? {};
-		const overrideProvider = overrideProviders[providerId] ?? {};
-		return [providerId, {
-			...baseProvider,
-			...overrideProvider
-		}];
-	}));
-	return {
-		...base,
-		...override,
-		modelOverrides: {
-			...base.modelOverrides,
-			...override.modelOverrides
-		},
-		...Object.keys(mergedProviders).length === 0 ? {} : { providers: mergedProviders }
-	};
-}
-function resolveVoiceTtsConfig(params) {
-	if (!params.override) return {
-		cfg: params.cfg,
-		resolved: resolveTtsConfig(params.cfg)
-	};
-	const merged = mergeTtsConfig(params.cfg.messages?.tts ?? {}, params.override);
-	const messages = params.cfg.messages ?? {};
-	const cfg = {
-		...params.cfg,
-		messages: {
-			...messages,
-			tts: merged
-		}
-	};
-	return {
-		cfg,
-		resolved: resolveTtsConfig(cfg)
-	};
-}
 async function transcribeVoiceAudio(params) {
 	return normalizeOptionalString((await getDiscordRuntime().mediaUnderstanding.transcribeAudioFile({
 		filePath: params.filePath,
@@ -2168,21 +2124,18 @@ async function transcribeVoiceAudio(params) {
 	})).text);
 }
 async function synthesizeVoiceReplyAudio(params) {
-	const { cfg: ttsCfg, resolved: ttsConfig } = resolveVoiceTtsConfig({
+	const runtime = getDiscordRuntime();
+	const prepared = await runtime.tts.prepareTtsRequest({
 		cfg: params.cfg,
-		override: params.override
-	});
-	const directive = parseTtsDirectives(params.replyText, ttsConfig.modelOverrides, {
-		cfg: ttsCfg,
-		providerConfigs: ttsConfig.providerConfigs,
-		preferredProviderId: getTtsProvider(ttsConfig, resolveTtsPrefsPath(ttsConfig))
+		override: params.override,
+		text: params.replyText
 	});
+	const directive = prepared.directives;
 	const speakText = sanitizeVoiceReplyTextForSpeech(directive.overrides.ttsText ?? directive.cleanedText.trim(), params.speakerLabel);
 	if (!speakText) return { status: "empty" };
-	const runtime = getDiscordRuntime();
 	const streamResult = await runtime.tts.textToSpeechStream?.({
 		text: speakText,
-		cfg: ttsCfg,
+		cfg: prepared.cfg,
 		channel: "discord",
 		overrides: directive.overrides,
 		disableFallback: true
@@ -2196,7 +2149,7 @@ async function synthesizeVoiceReplyAudio(params) {
 	};
 	const result = await runtime.tts.textToSpeech({
 		text: speakText,
-		cfg: ttsCfg,
+		cfg: prepared.cfg,
 		channel: "discord",
 		overrides: directive.overrides
 	});
@@ -2891,7 +2844,7 @@ var DiscordVoiceManager$1 = class {
 			};
 		}
 		const { route, voiceRoute, agentSessionMode, agentSessionTarget } = routeInfo;
-		logger.info(`discord voice: joining guild=${guildId} channel=${channelId} mode=${voiceMode} agent=${route.agentId} voiceSession=${voiceRoute.sessionKey} supervisorSession=${route.sessionKey} agentSessionMode=${agentSessionMode}${agentSessionTarget ? ` agentSessionTarget=${agentSessionTarget}` : ""} voiceModel=${voiceConfig?.model ?? "route-default"} realtimeProvider=${voiceConfig?.realtime?.provider ?? "auto"} realtimeModel=${voiceConfig?.realtime?.model ?? "provider-default"} realtimeVoice=${voiceConfig?.realtime?.voice ?? "provider-default"}`);
+		logger.info(`discord voice: joining guild=${guildId} channel=${channelId} mode=${voiceMode} agent=${route.agentId} voiceSession=${voiceRoute.sessionKey} supervisorSession=${route.sessionKey} agentSessionMode=${agentSessionMode}${agentSessionTarget ? ` agentSessionTarget=${agentSessionTarget}` : ""} voiceModel=${voiceConfig?.model ?? "route-default"} realtimeProvider=${voiceConfig?.realtime?.provider ?? "auto"} realtimeModel=${voiceConfig?.realtime?.model ?? "provider-default"} realtimeVoice=${voiceConfig?.realtime?.speakerVoice ?? voiceConfig?.realtime?.speakerVoiceId ?? "provider-default"}`);
 		const player = voiceSdk.createAudioPlayer();
 		connection.subscribe(player);
 		let stopped = false;