@xopcai/xopc 0.0.303 → 0.0.307

package/dist/src/agent/orchestration/agent-orchestrator.js

@@ -71,6 +71,7 @@ var AgentOrchestrator = class {
 			await this.sessionHydrator.model(conversationId);
 			const channelSystemPrompt = typeof context.metadata?.channelSystemPrompt === "string" ? context.metadata.channelSystemPrompt.trim() : "";
 			if (channelSystemPrompt) {
+				await this.agentManager.ensureMemoryReadyForSession?.(conversationId);
 				this.agentManager.getOrCreateAgent(conversationId);
 				this.agentManager.applyTurnChannelSystemPrompt(conversationId, channelSystemPrompt);
 			}
@@ -90,7 +91,8 @@ var AgentOrchestrator = class {
 			setPendingTranscriptUserMessage(conversationId, userMessage);
 			const userPlainForMemory = extractAgentUserPlainText(userMessage);
 			const turnId = randomUUID();
-			const userMessageForModel = (await this.agentManager.prepareUserTurnContext(userMessage, conversationId, turnId)).modelMessage;
+			const userContext = await this.agentManager.prepareUserTurnContext(userMessage, conversationId, turnId);
+			const userMessageForModel = userMessage;
 			const llmTurn = await hydrateUserTurnForLlm({
 				message: userMessage,
 				modelRef
@@ -101,6 +103,7 @@ var AgentOrchestrator = class {
 						conversationId,
 						runId: turnId,
 						userMessage: userMessageForModel,
+						dynamicSystemContext: userContext.dynamicSystemContext,
 						llmImages: llmTurn.images,
 						sessionStore: this.sessionStore,
 						agentManager: this.agentManager,

package/dist/src/agent/prompt/cache-boundary.d.ts

@@ -24,3 +24,5 @@ export declare function splitPromptCacheBoundary(text: string): {
 export declare function normalizePromptSection(text: string): string;
 /** Add stable content immediately before the cache boundary. */
 export declare function appendStablePromptSection(systemPrompt: string, section: string): string;
+/** Add per-turn context after the cache boundary so the stable prefix remains reusable. */
+export declare function appendDynamicPromptSection(systemPrompt: string, section: string): string;

package/dist/src/agent/prompt/cache-boundary.js

@@ -40,5 +40,14 @@ function appendStablePromptSection(systemPrompt, section) {
 	const stablePrefix = [split.stablePrefix, normalizedSection].filter(Boolean).join("\n\n");
 	return split.dynamicSuffix ? `${stablePrefix}${PROMPT_CACHE_BOUNDARY}${split.dynamicSuffix}` : `${stablePrefix}${PROMPT_CACHE_BOUNDARY}`;
 }
+/** Add per-turn context after the cache boundary so the stable prefix remains reusable. */
+function appendDynamicPromptSection(systemPrompt, section) {
+	const normalizedSection = normalizePromptSection(section);
+	if (!normalizedSection) return systemPrompt;
+	const split = splitPromptCacheBoundary(systemPrompt);
+	if (!split) return `${normalizePromptSection(systemPrompt)}${PROMPT_CACHE_BOUNDARY}${normalizedSection}`;
+	const dynamicSuffix = [split.dynamicSuffix, normalizedSection].filter(Boolean).join("\n\n");
+	return `${split.stablePrefix}${PROMPT_CACHE_BOUNDARY}${dynamicSuffix}`;
+}
 //#endregion
-export { PROMPT_CACHE_BOUNDARY, appendStablePromptSection, normalizePromptSection, splitPromptCacheBoundary, stripPromptCacheBoundary };
+export { PROMPT_CACHE_BOUNDARY, appendDynamicPromptSection, appendStablePromptSection, normalizePromptSection, splitPromptCacheBoundary, stripPromptCacheBoundary };

package/dist/src/agent/service/build-direct-message-content.d.ts

@@ -11,4 +11,5 @@ export declare function buildDirectUserMessageContent(opts: {
     config: Config;
     agentManager: AgentInstanceGateway;
     modelManager: ModelManager;
+    suppressMediaPromptUris?: ReadonlySet<string>;
 }): Promise<import("../inbound/attachment-pipeline.js").TranscriptUserMessage>;

package/dist/src/agent/service/build-direct-message-content.js

@@ -13,7 +13,8 @@ async function buildDirectUserMessageContent(opts) {
 		conversationId: sk,
 		modelRef: modelRef || getDefaultModelSync(opts.config),
 		config: opts.config,
-		agentManager: opts.agentManager
+		agentManager: opts.agentManager,
+		suppressMediaPromptUris: opts.suppressMediaPromptUris
 	});
 }
 //#endregion

package/dist/src/agent/service/direct-turn-helpers.js

@@ -3,7 +3,6 @@ import { getConnectionResumeInput, init_connection_wait_repository } from "../..
 import { resolveImageHandlingStrategy } from "../image/vision-detection.js";
 import { hydrateUserTurnForLlm } from "../inbound/attachment-pipeline.js";
 import { buildTaskExecutionDirective } from "../../tasks/task-context-assembler.js";
-import { prependAgentContext } from "../context/prepend.js";
 import { extractAgentUserPlainText } from "../memory/user-message-text.js";
 import { shouldSkipResetOverlapCommand } from "../../session/reset-triggers.js";
 import { parseSlashCommand } from "../../chat-commands/command-parse.js";
@@ -11,6 +10,7 @@ import { commandRegistry } from "../../chat-commands/registry.js";
 import "../../chat-commands/index.js";
 import { runEmbeddedTurnForSession } from "../embedded/run-for-session.js";
 import { injectSourceContextsIntoUserMessage } from "../source-context/injector.js";
+import { prependAgentContext } from "../context/prepend.js";
 import crypto from "node:crypto";
 //#region src/agent/service/direct-turn-helpers.ts
 /**
@@ -101,7 +101,7 @@ async function runDirectAgentTurn(deps, input) {
 	const userPlain = extractAgentUserPlainText(input.userMessage);
 	const userContext = await deps.agentManager.prepareUserTurnContext(input.userMessage, input.conversationId, turnId);
 	const sourceContexts = input.sourceContexts ?? [];
-	const userMessageForModel = prependAgentContext(injectSourceContextsIntoUserMessage(userContext.modelMessage, sourceContexts), buildTaskExecutionDirective(input.conversationId));
+	const userMessageForModel = prependAgentContext(injectSourceContextsIntoUserMessage(input.userMessage, sourceContexts), buildTaskExecutionDirective(input.conversationId));
 	const modelRef = deps.modelManager.getModelForSession(input.conversationId);
 	const llmTurn = await hydrateUserTurnForLlm({
 		message: input.userMessage,
@@ -114,6 +114,7 @@ async function runDirectAgentTurn(deps, input) {
 		conversationId: input.conversationId,
 		runId: turnId,
 		userMessage: userMessageForModel,
+		dynamicSystemContext: userContext.dynamicSystemContext,
 		llmImages,
 		presentation: input.presentation,
 		sessionStore: deps.sessionStore,

package/dist/src/agent/service/process-direct-streaming.d.ts

@@ -36,7 +36,9 @@ export interface ProcessDirectStreamingDeps {
     sessionConfigStore: SessionConfigStore;
     commandHandler: Pick<CommandHandler, 'executeCommandAndAggregateReply'>;
     prepareInboundAttachments: (conversationId: string, attachments?: DirectStreamInboundAttachment[]) => Promise<MediaRef[] | undefined>;
-    buildTranscriptUserMessage: (content: string, prepared: MediaRef[] | undefined, conversationId: string) => Promise<TranscriptUserMessage>;
+    buildTranscriptUserMessage: (content: string, prepared: MediaRef[] | undefined, conversationId: string, options?: {
+        suppressMediaPromptUris?: ReadonlySet<string>;
+    }) => Promise<TranscriptUserMessage>;
     recordTaskReviewStreamHint?: (conversationId: string, task: {
         skipTaskReview: boolean;
     }) => void;

package/dist/src/agent/service/process-direct-streaming.js

@@ -11,7 +11,7 @@ import { initSessionTurn } from "../../session/init-session-turn.js";
 import "../../session/index.js";
 import { mergeSttConfigFromAppConfig } from "../../voice/stt/config.js";
 import "../../voice/stt/index.js";
-import { isVoiceLikeAttachment, mergeVoiceTranscriptsIntoUserText } from "../../channels/attachments/voice-stt-webchat.js";
+import { isVoiceLikeAttachment, mergeVoiceTranscriptsIntoUserText, requestsOriginalVoiceInspection } from "../../channels/attachments/voice-stt-webchat.js";
 import { applyReasoningVisibility } from "../streaming/reasoning-visibility.js";
 import { isSessionSourceBinding, summarizeSourceContext } from "../source-context/types.js";
 import { injectSourceContextsIntoUserMessage } from "../source-context/injector.js";
@@ -207,6 +207,7 @@ async function* runProcessDirectStreaming(deps, input) {
 			const voiceMerge = await mergeVoiceTranscriptsIntoUserText(prepared, turnBody, sttCfg);
 			mergedUserText = voiceMerge.text;
 			inboundVoice = voiceMerge.inboundVoice;
+			const suppressMediaPromptUris = requestsOriginalVoiceInspection(mergedUserText) ? void 0 : new Set(voiceMerge.transcribedMediaUris ?? []);
 			if (inboundVoice) {
 				const transcriptParts = [voiceMerge.voiceTranscripts.filter(Boolean).join("\n"), turnBody.trim()].filter(Boolean);
 				const voiceMedia = (prepared ?? []).filter(isVoiceLikeAttachment).map((att) => ({
@@ -275,7 +276,7 @@ async function* runProcessDirectStreaming(deps, input) {
 			}
 			const skillTurn = deps.agentManager.prepareSkillTurn(conversationId, mergedUserText);
 			const textForAgent = skillTurn.text;
-			const userMessage = await deps.buildTranscriptUserMessage(textForAgent, prepared, conversationId);
+			const userMessage = await deps.buildTranscriptUserMessage(textForAgent, prepared, conversationId, { suppressMediaPromptUris });
 			const sourceContexts = [...input.sourceContexts ?? []];
 			if (deps.sourceContextResolver) {
 				const metadata = await deps.sessionStore.getMetadata(conversationId).catch(() => null);

package/dist/src/agent/service/webchat-tts.d.ts

@@ -1,6 +1,4 @@
 import type { Config } from '../../config/schema.js';
-import type { MediaRef } from '../../media/types.js';
-import type { SessionStore } from '../../session/index.js';
 export type WebchatTtsResult = {
     type: 'tts_audio';
     uri: string;
@@ -9,7 +7,6 @@ export type WebchatTtsResult = {
 };
 export type WebchatTtsDeps = {
     config: Config | undefined;
-    sessionStore: SessionStore;
     getLastAssistantPlainText: (conversationId: string) => string;
     log: {
         warn: (obj: Record<string, unknown>, msg: string) => void;
@@ -21,4 +18,3 @@ export declare function isSuccessfulWebchatTtsToolEvent(event: unknown): boolean
  * Generate TTS for webchat when config allows; persist under `{stateDir}/media/tts/`.
  */
 export declare function maybeEmitWebchatTts(deps: WebchatTtsDeps, conversationId: string, hadInboundVoice: boolean): Promise<WebchatTtsResult | null>;
-export declare function appendMediaToLastAssistant(sessionStore: SessionStore, conversationId: string, ref: MediaRef): Promise<void>;

package/dist/src/agent/service/webchat-tts.js

@@ -1,3 +1,5 @@
+import { appendMediaToAssistantTranscriptEntry, findLatestAssistantTranscriptEntryId } from "../../storage/sqlite/transcript-repository.js";
+import "../../storage/sqlite/index.js";
 import { isTTSAvailable } from "../../voice/tts/factory.js";
 import { speak } from "../../voice/tts/speak-core.js";
 import { mergeTtsConfigFromAppConfig } from "../../voice/tts/merge-config.js";
@@ -5,6 +7,7 @@ import { getChannelOutputFormat, shouldUseTTS } from "../../voice/tts/service.js
 import { compressAudio } from "../../voice/tts/audio.js";
 import { persistOutboundTtsAudio } from "../../channels/attachments/outbound-tts-persist.js";
 import "../../voice/tts/index.js";
+import { deleteMediaUris } from "../../media/session-references.js";
 //#region src/agent/service/webchat-tts.ts
 /** True when the explicit TTS tool already produced durable Webchat audio. */
 function isSuccessfulWebchatTtsToolEvent(event) {
@@ -28,6 +31,12 @@ async function maybeEmitWebchatTts(deps, conversationId, hadInboundVoice) {
 	if (!shouldUseTTS(ttsConfig, hadInboundVoice).useTTS) return null;
 	const text = deps.getLastAssistantPlainText(conversationId).trim();
 	if (!text) return null;
+	const assistantEntryId = findLatestAssistantTranscriptEntryId(conversationId);
+	if (!assistantEntryId) {
+		deps.log.warn({ conversationId }, "Webchat TTS skipped because the assistant transcript is unavailable");
+		return null;
+	}
+	let persisted;
 	try {
 		const webOut = getChannelOutputFormat("webchat");
 		const fmt = webOut.format;
@@ -37,8 +46,15 @@ async function maybeEmitWebchatTts(deps, conversationId, hadInboundVoice) {
 		});
 		const { buffer, format } = await compressAudio(Buffer.from(ttsResult.audio), ttsResult.format, webOut.format === "mp3" ? "mp3" : "opus");
 		const normalizedMime = format === "opus" || format === "ogg" ? "audio/ogg" : format === "mp3" || format === "mpeg" ? "audio/mpeg" : format === "wav" ? "audio/wav" : `audio/${format}`;
-		const persisted = await persistOutboundTtsAudio(buffer, format);
-		await appendMediaToLastAssistant(deps.sessionStore, conversationId, persisted);
+		persisted = await persistOutboundTtsAudio(buffer, format);
+		if (!appendMediaToAssistantTranscriptEntry(conversationId, assistantEntryId, persisted)) {
+			await deleteMediaUris([persisted.uri]);
+			deps.log.warn({
+				conversationId,
+				assistantEntryId
+			}, "Webchat TTS discarded because its assistant transcript changed");
+			return null;
+		}
 		return {
 			type: "tts_audio",
 			uri: persisted.uri,
@@ -46,6 +62,7 @@ async function maybeEmitWebchatTts(deps, conversationId, hadInboundVoice) {
 			name: persisted.name
 		};
 	} catch (err) {
+		if (persisted) await deleteMediaUris([persisted.uri]);
 		deps.log.warn({
 			err,
 			conversationId
@@ -53,21 +70,5 @@ async function maybeEmitWebchatTts(deps, conversationId, hadInboundVoice) {
 		return null;
 	}
 }
-async function appendMediaToLastAssistant(sessionStore, conversationId, ref) {
-	const loaded = await sessionStore.load(conversationId);
-	for (let i = loaded.length - 1; i >= 0; i--) {
-		const m = loaded[i];
-		if (m.role === "assistant") {
-			const prev = m.media ?? [];
-			if (prev.some((x) => x.uri === ref.uri)) return;
-			loaded[i] = {
-				...m,
-				media: [...prev, ref]
-			};
-			await sessionStore.saveMessages(conversationId, loaded);
-			return;
-		}
-	}
-}
 //#endregion
-export { appendMediaToLastAssistant, isSuccessfulWebchatTtsToolEvent, maybeEmitWebchatTts };
+export { isSuccessfulWebchatTtsToolEvent, maybeEmitWebchatTts };

package/dist/src/agent/tools/factory.js

@@ -389,7 +389,7 @@ var AgentToolsFactory = class AgentToolsFactory {
 				canRead: () => currentAccess().knowledge,
 				canWrite: () => currentAccess().knowledge,
 				getWritePolicy: knowledgeWritePolicy,
-				getSources: () => currentAccess().knowledgeSources
+				getReadPolicy: () => currentAccess().knowledgePolicy
 			}),
 			createKnowledgeGetTool({
 				agentId: resolvedAgentId,
@@ -399,7 +399,7 @@ var AgentToolsFactory = class AgentToolsFactory {
 				canRead: () => currentAccess().knowledge,
 				canWrite: () => currentAccess().knowledge,
 				getWritePolicy: knowledgeWritePolicy,
-				getSources: () => currentAccess().knowledgeSources
+				getReadPolicy: () => currentAccess().knowledgePolicy
 			}),
 			createKnowledgeWriteTool({
 				agentId: resolvedAgentId,
@@ -409,7 +409,7 @@ var AgentToolsFactory = class AgentToolsFactory {
 				canRead: () => currentAccess().knowledge,
 				canWrite: () => currentAccess().knowledge,
 				getWritePolicy: knowledgeWritePolicy,
-				getSources: () => currentAccess().knowledgeSources
+				getReadPolicy: () => currentAccess().knowledgePolicy
 			}),
 			...this.deps.getSessionStore ? [createSessionRecallTool({
 				getSessionStore: this.deps.getSessionStore,

package/dist/src/agent/tools/knowledge-memory-tool.d.ts

@@ -1,5 +1,5 @@
 import type { AgentTool } from '@earendil-works/pi-agent-core';
-import { type KnowledgeSource } from '../../knowledge-memory/index.js';
+import { type KnowledgeReadPolicy } from '../../knowledge-memory/index.js';
 export interface KnowledgeToolOptions {
     agentId: string;
     workspaceId: string;
@@ -8,7 +8,7 @@ export interface KnowledgeToolOptions {
     canRead: () => boolean;
     canWrite: () => boolean;
     getWritePolicy: () => 'deny' | 'confirm' | 'allow';
-    getSources: () => readonly KnowledgeSource[];
+    getReadPolicy: () => KnowledgeReadPolicy;
 }
 export declare function createKnowledgeSearchTool(options: KnowledgeToolOptions): AgentTool;
 export declare function createKnowledgeGetTool(options: KnowledgeToolOptions): AgentTool;

package/dist/src/agent/tools/knowledge-memory-tool.js

@@ -1,4 +1,4 @@
-import { getKnowledgeItem, knowledgeSourceAllowed, searchKnowledgeItems, writeKnowledgeItem } from "../../knowledge-memory/repository.js";
+import { getKnowledgeItem, knowledgeItemAllowed, searchKnowledgeItems, writeKnowledgeItem } from "../../knowledge-memory/repository.js";
 import "../../knowledge-memory/index.js";
 import { Type } from "@sinclair/typebox";
 //#region src/agent/tools/knowledge-memory-tool.ts
@@ -60,7 +60,7 @@ function createKnowledgeSearchTool(options) {
 			const results = searchKnowledgeItems({
 				query: input.query,
 				context: visibility(options),
-				sources: options.getSources(),
+				policy: options.getReadPolicy(),
 				limit: input.maxResults ?? 12
 			});
 			return {
@@ -92,7 +92,7 @@ function createKnowledgeGetTool(options) {
 			};
 			const id = raw.id;
 			const item = getKnowledgeItem(id);
-			return item && visibleItem(options, item) && knowledgeSourceAllowed(item, options.getSources()) ? {
+			return item && visibleItem(options, item) && knowledgeItemAllowed(item, options.getReadPolicy()) ? {
 				content: [{
 					type: "text",
 					text: JSON.stringify(item, null, 2)
@@ -131,14 +131,14 @@ function createKnowledgeWriteTool(options) {
 				details: { error: "knowledge_write_denied" }
 			};
 			const input = raw;
-			if (!options.getSources().includes(input.scope)) return {
+			if (!options.getReadPolicy().scopes.includes(input.scope) || !options.getReadPolicy().contentSources.includes("memory")) return {
 				content: [{
 					type: "text",
-					text: `Knowledge source is disabled: ${input.scope}`
+					text: `Knowledge memory writes are disabled for scope: ${input.scope}`
 				}],
 				details: {
-					error: "knowledge_source_disabled",
-					source: input.scope
+					error: "knowledge_scope_disabled",
+					scope: input.scope
 				}
 			};
 			const sessionId = options.getSessionId();

package/dist/src/browser/native-messaging-host.js

@@ -1,5 +1,5 @@
 import { closeXopcDatabase, init_connection, openXopcDatabase } from "../storage/sqlite/connection.js";
-import { devicePairingInvitationPayloadSchema, formatBrowserPairingInvitation } from "../../packages/gateway-contract/src/device-pairing.js";
+import { browserPairingInvitationPayloadSchema, formatBrowserPairingInvitation } from "../../packages/gateway-contract/src/device-pairing.js";
 import { init_src } from "../../packages/gateway-contract/src/index.js";
 import { getGatewayIdentityPublicKeyRaw, getOrCreateGatewayIdentity } from "../storage/sqlite/gateway-identity-repository.js";
 import { createDevicePairingSetup } from "../storage/sqlite/device-pairing-repository.js";
@@ -53,7 +53,7 @@ function createBrowserNativeBootstrap(configPath, request) {
 		return {
 			ok: true,
 			gatewayUrl,
-			invitation: formatBrowserPairingInvitation(Buffer.from(JSON.stringify(devicePairingInvitationPayloadSchema.parse({
+			invitation: formatBrowserPairingInvitation(Buffer.from(JSON.stringify(browserPairingInvitationPayloadSchema.parse({
 				version: 3,
 				pairingToken: setup.token,
 				gatewayId: identity.id,

package/dist/src/channels/attachments/voice-stt-webchat.d.ts

@@ -4,10 +4,13 @@
 import type { STTConfig } from '../../voice/stt/types.js';
 import { type InboundAttachmentInput, type MediaRef } from './inbound-persist.js';
 export declare function isVoiceLikeAttachment(att: InboundAttachmentInput | MediaRef): boolean;
+/** True when the user explicitly asks the agent to inspect the original recording. */
+export declare function requestsOriginalVoiceInspection(text: string): boolean;
 export declare function mergeVoiceTranscriptsIntoUserText(prepared: (InboundAttachmentInput | MediaRef)[] | undefined, userText: string, sttConfig: STTConfig, opts?: {
     skipVoiceTranscription?: boolean;
 }): Promise<{
     text: string;
     inboundVoice: boolean;
     voiceTranscripts: string[];
+    transcribedMediaUris: string[];
 }>;

package/dist/src/channels/attachments/voice-stt-webchat.js

@@ -12,28 +12,39 @@ function isVoiceLikeAttachment(att) {
 	if (att.type === "voice") return true;
 	return (att.mimeType?.toLowerCase() ?? "").startsWith("audio/");
 }
+/** True when the user explicitly asks the agent to inspect the original recording. */
+function requestsOriginalVoiceInspection(text) {
+	const normalized = text.trim();
+	if (!normalized) return false;
+	return /(?:分析|检查|听|辨别|识别|判断|检测|对比|提取).{0,12}(?:原始|原声)?(?:语音|音频|录音|声音)/u.test(normalized) || /(?:原始|原声)?(?:语音|音频|录音|声音).{0,12}(?:分析|检查|听|辨别|识别|判断|检测|对比|音质|音色|语气|情绪|口音|噪声|杂音|背景声|说话人|声纹)/u.test(normalized) || /\b(?:analy[sz]e|inspect|listen\s+to|check|identify|compare|detect)\b.{0,40}\b(?:(?:original|raw)\s+)?(?:audio|recording|voice(?:\s+(?:clip|message))?)\b/iu.test(normalized) || /\b(?:(?:original|raw)\s+)?(?:audio|recording|voice(?:\s+(?:clip|message))?)\b.{0,40}\b(?:analy[sz]e|inspect|listen|quality|tone|emotion|accent|noise|background|speaker|voiceprint)\b/iu.test(normalized);
+}
 async function mergeVoiceTranscriptsIntoUserText(prepared, userText, sttConfig, opts) {
 	if (!prepared?.length) return {
 		text: userText,
 		inboundVoice: false,
-		voiceTranscripts: []
+		voiceTranscripts: [],
+		transcribedMediaUris: []
 	};
 	if (!prepared.some(isVoiceLikeAttachment)) return {
 		text: userText,
 		inboundVoice: false,
-		voiceTranscripts: []
+		voiceTranscripts: [],
+		transcribedMediaUris: []
 	};
 	if (opts?.skipVoiceTranscription === true) return {
 		text: userText,
 		inboundVoice: true,
-		voiceTranscripts: []
+		voiceTranscripts: [],
+		transcribedMediaUris: []
 	};
 	if (!isSTTAvailable(sttConfig)) return {
 		text: userText,
 		inboundVoice: true,
-		voiceTranscripts: []
+		voiceTranscripts: [],
+		transcribedMediaUris: []
 	};
 	const transcripts = [];
+	const transcribedMediaUris = [];
 	for (const att of prepared) {
 		if (!isVoiceLikeAttachment(att)) continue;
 		let buf = null;
@@ -56,12 +67,13 @@ async function mergeVoiceTranscriptsIntoUserText(prepared, userText, sttConfig,
 			continue;
 		}
 		try {
-			const r = await transcribe(buf, sttConfig, {
+			const transcript = (await transcribe(buf, sttConfig, {
 				language: sttConfig.provider === "alibaba" ? "zh" : void 0,
 				mime: att.mimeType,
 				fileName: att.name
-			});
-			transcripts.push(r.text.trim() || "[Voice: no speech detected]");
+			})).text.trim();
+			transcripts.push(transcript || "[Voice: no speech detected]");
+			if (transcript && "uri" in att && att.uri?.trim()) transcribedMediaUris.push(att.uri.trim());
 		} catch (error) {
 			const errorMessage = error instanceof Error ? error.message : String(error);
 			log.error({
@@ -78,8 +90,9 @@ async function mergeVoiceTranscriptsIntoUserText(prepared, userText, sttConfig,
 	return {
 		text: [transcripts.filter(Boolean).join("\n"), userText.trim()].filter(Boolean).join("\n\n") || userText,
 		inboundVoice: true,
-		voiceTranscripts: transcripts
+		voiceTranscripts: transcripts,
+		transcribedMediaUris
 	};
 }
 //#endregion
-export { isVoiceLikeAttachment, mergeVoiceTranscriptsIntoUserText };
+export { isVoiceLikeAttachment, mergeVoiceTranscriptsIntoUserText, requestsOriginalVoiceInspection };

package/dist/src/cli/commands/agent.js

@@ -77,6 +77,7 @@ function createAgentCommand(_ctx) {
 			extensionLoader.setConfig(config);
 			extensionLoader.setRuntimeContext({ bus });
 			await extensionLoader.loadByActivationPlan();
+			await extensionLoader.startServices();
 			const n = extensionLoader.getRegistry().extensions.size;
 			if (n > 0) log.info({ count: n }, "Extensions loaded");
 		} catch (error) {
@@ -120,6 +121,7 @@ function createAgentCommand(_ctx) {
 			running = false;
 			bus.shutdown();
 			await agent.stop();
+			await extensionLoader?.shutdown();
 		};
 		process.on("SIGINT", shutdown);
 		process.on("SIGTERM", shutdown);

package/dist/src/cli/commands/config.js

@@ -1,17 +1,16 @@
 import { ConfigSchema, init_schema } from "../../config/schema.js";
-import { init_write_file_atomic, writeTextAtomic } from "../../infra/write-file-atomic.js";
 import { resolveGatewayLocalClientHost } from "../../config/gateway-bind.js";
 import { formatExamples, register } from "../registry.js";
 import { existsSync, readFileSync } from "fs";
 import { Command } from "commander";
 //#region src/cli/commands/config.ts
-init_write_file_atomic();
 init_schema();
 const MISSING_CONFIG_HINT = "Run: xopc setup, xopc onboard, or xopc init";
 async function loadConfigDeps() {
-	const [{ loadConfig }, { createLogger }] = await Promise.all([import("../../config/index.js"), import("../../utils/logger.js")]);
+	const [{ loadConfig, saveConfig }, { createLogger }] = await Promise.all([import("../../config/index.js"), import("../../utils/logger.js")]);
 	return {
 		loadConfig,
+		saveConfig,
 		log: createLogger("ConfigCommand")
 	};
 }
@@ -108,7 +107,7 @@ function createConfigCommand(ctx) {
 		console.log(typeof value === "object" ? JSON.stringify(value, null, 2) : value);
 	});
 	cmd.command("set <path> <value>").description("Set a config value by dot path").action(async (path, value) => {
-		const { loadConfig, log } = await loadConfigDeps();
+		const { loadConfig, saveConfig, log } = await loadConfigDeps();
 		if (!existsSync(ctx.configPath)) {
 			log.error(`Config file not found. ${MISSING_CONFIG_HINT}`);
 			process.exit(1);
@@ -121,11 +120,11 @@ function createConfigCommand(ctx) {
 		}
 		const config = loadConfig(ctx.configPath);
 		setNestedValue(config, path, parsedValue);
-		await writeTextAtomic(ctx.configPath, JSON.stringify(config, null, 2));
+		await saveConfig(config, ctx.configPath);
 		log.info({ path }, `Config updated`);
 	});
 	cmd.command("unset <path>").description("Remove a config value by dot path").action(async (path) => {
-		const { loadConfig, log } = await loadConfigDeps();
+		const { loadConfig, saveConfig, log } = await loadConfigDeps();
 		if (!existsSync(ctx.configPath)) {
 			log.error(`Config file not found. ${MISSING_CONFIG_HINT}`);
 			process.exit(1);
@@ -138,7 +137,7 @@ function createConfigCommand(ctx) {
 		}, config);
 		if (target && typeof target === "object" && lastKey in target) {
 			delete target[lastKey];
-			await writeTextAtomic(ctx.configPath, JSON.stringify(config, null, 2));
+			await saveConfig(config, ctx.configPath);
 			log.info({ path }, `Config removed`);
 		} else {
 			log.error({ path }, `Config path not found`);
@@ -152,7 +151,7 @@ function createConfigCommand(ctx) {
 		if (!await runConfigValidate(ctx.configPath)) process.exit(1);
 	});
 	cmd.command("token").description("Generate or view the gateway auth token").option("--generate", "Generate a new token").option("--show", "Show the current token (unmasked)").action(async (options) => {
-		const { loadConfig, log } = await loadConfigDeps();
+		const { loadConfig, saveConfig, log } = await loadConfigDeps();
 		if (!existsSync(ctx.configPath)) {
 			log.error(`Config file not found. ${MISSING_CONFIG_HINT}`);
 			process.exit(1);
@@ -171,7 +170,7 @@ function createConfigCommand(ctx) {
 				mode: "token",
 				token: newToken
 			};
-			await writeTextAtomic(ctx.configPath, JSON.stringify(config, null, 2));
+			await saveConfig(config, ctx.configPath);
 			log.info("New gateway token generated");
 			console.log(`Token: ${newToken.slice(0, 8)}...${newToken.slice(-8)}`);
 			console.log("\nUse \"xopc config token --show\" to view the full token");

package/dist/src/cli/commands/doctor/checks/database-schema.js

@@ -53,7 +53,7 @@ async function checkDatabaseSchema(ctx) {
 		id: "database-schema",
 		label: "Database schema",
 		status: "pass",
-		message: `SQLite schema v${status.dbVersion} (current release v179).`,
+		message: `SQLite schema v${status.dbVersion} (current release v183).`,
 		hints: []
 	};
 }