@xopcai/xopc 0.0.364 → 0.0.366

package/dist/src/voice/realtime/agentEngine.js

@@ -39,10 +39,13 @@ function createAgentVoiceEngine(options) {
 	const echoCandidates = new PlaybackEchoCandidates();
 	const finalizedUtterances = /* @__PURE__ */ new Set();
 	let pendingTurn;
+	let turnReceivedAtMs;
 	let taskUpdateQueued = false;
 	let lastUserSpeechAt = 0;
 	const turn = new TurnCoordinator(claim.silenceDurationMs, (text, decision, turnId) => {
 		if (closed || muted) return;
+		const receivedAtMs = turnReceivedAtMs ?? Date.now();
+		turnReceivedAtMs = void 0;
 		log.info({
 			sessionId: claim.sessionId,
 			turnId,
@@ -77,12 +80,14 @@ function createAgentVoiceEngine(options) {
 		const pending = {
 			turnId,
 			text,
-			cancelled: false
+			cancelled: false,
+			receivedAtMs,
+			committedAtMs: Date.now()
 		};
 		pendingTurn = pending;
 		conversationTail = conversationTail.then(() => {
 			if (pendingTurn === pending) pendingTurn = void 0;
-			if (generation === inputGeneration && !pending.cancelled) return runAssistantTurn(text, turnId).then(() => {});
+			if (generation === inputGeneration && !pending.cancelled) return runAssistantTurn(text, turnId, void 0, pending.receivedAtMs, pending.committedAtMs).then(() => {});
 		}).finally(() => {
 			if (generation === inputGeneration) queuedTurns -= 1;
 		});
@@ -156,7 +161,11 @@ function createAgentVoiceEngine(options) {
 					log.info({
 						sessionId: claim.sessionId,
 						responseId: response.id,
-						latencyMs: Date.now() - response.startedAt
+						conversationId: claim.request.conversationId,
+						phase: "first_response_audio",
+						latencyMs: Date.now() - response.startedAt,
+						inputToAudioMs: Date.now() - response.inputReceivedAtMs,
+						textToAudioMs: response.firstTextAtMs === void 0 ? void 0 : Date.now() - response.firstTextAtMs
 					}, "Realtime voice first audio ready");
 					send("response.audio.started", {
 						responseId: response.id,
@@ -241,7 +250,7 @@ function createAgentVoiceEngine(options) {
 		if (response.speechWorker) prefetchNextSpeech(response);
 		else startSpeechWorker(response);
 	}
-	async function runAssistantTurn(text, turnId, updateClientMessageId) {
+	async function runAssistantTurn(text, turnId, updateClientMessageId, receivedAtMs = Date.now(), committedAtMs = receivedAtMs) {
 		if (!claim.tts || !claim.request.conversationId || !text.trim() || closed) return false;
 		cancelActiveResponse("barge_in");
 		const response = {
@@ -255,6 +264,8 @@ function createAgentVoiceEngine(options) {
 			speechQueue: [],
 			queuedSpeechCharacters: 0,
 			startedAt: Date.now(),
+			inputReceivedAtMs: receivedAtMs,
+			turnCommittedAtMs: committedAtMs,
 			firstTextSeen: false,
 			awaitingClarification: false,
 			taskDone: false,
@@ -289,12 +300,18 @@ function createAgentVoiceEngine(options) {
 				if (event.type === "assistant_delta" || event.type === "tool_end" && event.payload?.toolName !== "clarify") response.awaitingClarification = false;
 				if (event.type === "assistant_delta" && typeof event.payload?.delta === "string") {
 					if (response.text.length + event.payload.delta.length > 32e3) throw new Error("Voice response text limit reached");
-					if (!response.firstTextSeen) {
+					if (!response.firstTextSeen && event.payload.delta.trim()) {
 						response.firstTextSeen = true;
+						response.firstTextAtMs = Date.now();
 						log.info({
 							sessionId: claim.sessionId,
 							responseId: response.id,
-							latencyMs: Date.now() - response.startedAt
+							conversationId: claim.request.conversationId,
+							phase: "first_response_text",
+							latencyMs: Date.now() - response.startedAt,
+							inputToTextMs: Date.now() - response.inputReceivedAtMs,
+							turnDecisionMs: Math.max(0, response.turnCommittedAtMs - response.inputReceivedAtMs),
+							queueMs: Math.max(0, response.startedAt - response.turnCommittedAtMs)
 						}, "Realtime voice first response text ready");
 					}
 					response.text += event.payload.delta;
@@ -422,6 +439,7 @@ function createAgentVoiceEngine(options) {
 		});
 	}
 	function bufferFinal(utteranceId, text) {
+		if (text && (turnReceivedAtMs === void 0 || turn.isIdle())) turnReceivedAtMs = Date.now();
 		try {
 			turn.final(utteranceId, text);
 		} catch (error) {
@@ -464,6 +482,7 @@ function createAgentVoiceEngine(options) {
 				if (pendingTurn) {
 					pendingTurn.cancelled = true;
 					turn.restore(pendingTurn.turnId, pendingTurn.text);
+					turnReceivedAtMs = pendingTurn.receivedAtMs;
 					pendingTurn = void 0;
 				}
 				turn.start(event.utteranceId);
@@ -512,6 +531,7 @@ function createAgentVoiceEngine(options) {
 				...event.language === "zh" || event.language === "en" ? { language: event.language } : {}
 			});
 			if (claim.request.purpose === "conversation") {
+				if (claim.request.conversationId) options.runtime.onConversationInput?.(claim.request.conversationId);
 				if (activeResponse?.awaitingClarification) {
 					turn.reset();
 					return;
@@ -525,6 +545,7 @@ function createAgentVoiceEngine(options) {
 		turn.reset();
 		echoCandidates.clear();
 		pendingTurn = void 0;
+		turnReceivedAtMs = void 0;
 		inputGeneration += 1;
 		queuedTurns = 0;
 		bufferedAudio = [];
@@ -548,10 +569,14 @@ function createAgentVoiceEngine(options) {
 		});
 		return inputReset;
 	}
+	function isConversationIdle() {
+		return claim.request.purpose === "conversation" && !closed && Boolean(sttSession) && !committing && !activeResponse && !pendingTurn && queuedTurns === 0 && turn.isIdle() && Date.now() - lastUserSpeechAt >= TASK_UPDATE_QUIET_MS;
+	}
 	function canOfferTaskUpdateInput() {
-		return claim.request.purpose === "conversation" && Boolean(claim.tts) && !closed && Boolean(sttSession) && !committing && !activeResponse && !pendingTurn && queuedTurns === 0 && turn.isIdle() && Date.now() - lastUserSpeechAt >= TASK_UPDATE_QUIET_MS;
+		return Boolean(claim.tts) && isConversationIdle();
 	}
 	return {
+		isConversationIdle,
 		canOfferTaskUpdate() {
 			return !taskUpdateQueued && canOfferTaskUpdateInput();
 		},

package/dist/src/voice/realtime/engine.d.ts

@@ -10,6 +10,7 @@ export interface VoiceEngine {
     cancel(responseId: string, reason: 'client_cancelled' | 'barge_in'): boolean;
     cancelTask(taskId: string): Promise<boolean>;
     acknowledge(responseId: string, playedDurationMs: number): void;
+    isConversationIdle?(): boolean;
     canOfferTaskUpdate?(): boolean;
     offerTaskUpdate?(input: {
         clientMessageId: string;

package/dist/src/voice/realtime/omniEngine.d.ts

@@ -13,6 +13,7 @@ export declare function createOmniVoiceEngine(options: {
     bargeIn: boolean;
     send: VoiceEventSink;
     sendAudio: (responseId: string, bytes: Uint8Array) => void;
+    onUserInput?: () => void;
     record: (entry: OmniTranscript) => Promise<void>;
     onClose: (reason: string, notify: boolean) => Promise<void>;
 }): VoiceEngine;

package/dist/src/voice/realtime/omniEngine.js

@@ -44,6 +44,7 @@ function createOmniVoiceEngine(options) {
 	const pendingInputs = /* @__PURE__ */ new Set();
 	let failed = false;
 	let platformRequestId;
+	let lastInputAt = 0;
 	let active;
 	const echoCandidates = new PlaybackEchoCandidates();
 	let recentPlayback;
@@ -360,6 +361,7 @@ function createOmniVoiceEngine(options) {
 							clearingInput = false;
 							clearTimeout(clearTimer);
 						} else if (event.type === "input_audio_buffer.speech_started") {
+							lastInputAt = Date.now();
 							if (muted || clearingInput) {
 								discardedInputs.add(String(event.item_id));
 								return;
@@ -375,6 +377,7 @@ function createOmniVoiceEngine(options) {
 							if (active && !active.published) cancel("barge_in");
 							options.send("input.speech_started", { utteranceId: String(event.item_id) });
 						} else if (event.type === "input_audio_buffer.speech_stopped") {
+							lastInputAt = Date.now();
 							if (muted || inputBlocked || discardedInputs.has(String(event.item_id)) || recorded.has(String(event.item_id))) return;
 							speaking.delete(String(event.item_id));
 							turn.stop(String(event.item_id));
@@ -408,6 +411,10 @@ function createOmniVoiceEngine(options) {
 								cancel("barge_in");
 							}
 							turn.final(event.item_id, text);
+							if (text) {
+								lastInputAt = Date.now();
+								options.onUserInput?.();
+							}
 							save({
 								itemId: event.item_id,
 								role: "user",
@@ -567,6 +574,9 @@ function createOmniVoiceEngine(options) {
 			}
 			pumpInput();
 		},
+		isConversationIdle() {
+			return ready && !closed && !failed && !active && !speaking.size && !pendingInputs.size && !responseCreateQueued && !waitingForCancellation && !clearingInput && !inputBlocked && Date.now() - lastInputAt >= 750;
+		},
 		async commit() {
 			throw new Error("Natural conversation uses automatic turn detection");
 		},

package/dist/src/voice/realtime/persona-context.js

@@ -1,6 +1,6 @@
 import { init_agent_scope, resolveAgentProfileDir } from "../../agent/agent-scope.js";
-import { DEFAULT_SOUL_FILENAME } from "../../agent/context/workspace.js";
 import { resolveEffectiveAgentProfileForSession } from "../../config/agent-profile.js";
+import { DEFAULT_SOUL_FILENAME } from "../../agent/context/workspace.js";
 import { loadProfileBootstrapFiles } from "../../agent/bootstrap/load-bootstrap-files.js";
 import { join } from "node:path";
 import { statSync } from "node:fs";

package/dist/src/voice/realtime/runtime.d.ts

@@ -27,6 +27,7 @@ export declare class VoiceRealtimeRuntime {
     createSession(request: CreateVoiceSessionRequest, principalId: string, now?: number): Promise<CreateVoiceSessionResponse>;
     handleUpgrade(req: IncomingMessage, socket: Socket, head: Buffer): boolean;
     hasConversation(conversationId: string): boolean;
+    canPublishReply(conversationId: string): boolean;
     canOfferTaskUpdate(conversationId: string): boolean;
     offerTaskUpdate(input: {
         conversationId: string;

package/dist/src/voice/realtime/runtime.js

@@ -313,6 +313,10 @@ var VoiceRealtimeRuntime = class {
 		this.pruneTickets(Date.now());
 		return this.conversationReservations.has(conversationId);
 	}
+	canPublishReply(conversationId) {
+		const sessionId = this.conversationReservations.get(conversationId);
+		return (sessionId ? this.activeSessions.get(sessionId) : void 0)?.engine?.isConversationIdle?.() === true;
+	}
 	canOfferTaskUpdate(conversationId) {
 		const sessionId = this.conversationReservations.get(conversationId);
 		return (sessionId ? this.activeSessions.get(sessionId) : void 0)?.engine?.canOfferTaskUpdate?.() === true;
@@ -539,6 +543,7 @@ var VoiceRealtimeRuntime = class {
 					bargeIn: consumed.config.voice?.realtime?.bargeIn ?? true,
 					send,
 					sendAudio,
+					onUserInput: () => this.options.onConversationInput?.(consumed.request.conversationId),
 					record: (entry) => this.options.recordOmniTranscript(consumed.request.conversationId, consumed.sessionId, entry, consumed.conversationSessionId),
 					onClose: shutdown
 				}) : createAgentVoiceEngine({

package/dist/src/voice/realtime/runtime.types.d.ts

@@ -41,6 +41,8 @@ export interface ResolvedStreamingTts {
     route: VoiceProviderRoute;
 }
 export interface VoiceRealtimeRuntimeOptions {
+    /** Confirmed non-echo speech invalidates background reply drafts before turn commit. */
+    onConversationInput?: (conversationId: string) => void;
     isPrincipalActive?: (principalId: string) => boolean;
     recordOmniTranscript?: (conversationId: string, callId: string, entry: OmniTranscript, expectedTranscriptId: string) => Promise<void>;
     getConversationContext?: (conversationId: string, expectedTranscriptId: string) => Promise<VoiceConversationContext>;

package/dist/src/work-discovery/service.js

@@ -12,10 +12,10 @@ import { resolveEffectiveAgentProfile } from "../config/agent-profile.js";
 import { init_providers, resolveModel } from "../providers/index.js";
 import { writeKnowledgeItem } from "../knowledge-memory/repository.js";
 import "../knowledge-memory/index.js";
-import { createWorkDiscoveryRun, getProjectUnderstandingRun, getWorkDiscoveryOnboardingState, getWorkDiscoveryRun, getWorkDiscoveryRunByIdempotencyKey, listPendingProjectUnderstandingRuns, setWorkDiscoveryFeedback, setWorkDiscoveryOnboardingState, updateWorkDiscoveryRun } from "./repository.js";
 import { linkAssertionEvidence, reconcileAssertion } from "../user-model/repository.js";
 import { createPriorityWindow, createUserGoal } from "../user-model/goals.js";
 import "../user-model/index.js";
+import { createWorkDiscoveryRun, getProjectUnderstandingRun, getWorkDiscoveryOnboardingState, getWorkDiscoveryRun, getWorkDiscoveryRunByIdempotencyKey, listPendingProjectUnderstandingRuns, setWorkDiscoveryFeedback, setWorkDiscoveryOnboardingState, updateWorkDiscoveryRun } from "./repository.js";
 import { isLocalModelBaseUrl } from "../providers/model-call.js";
 import { createUnderstandingSourceRun, getUnderstandingSourceGrant, getUnderstandingSourceRun, updateUnderstandingSourceGrantCheckpoint, updateUnderstandingSourceGrantPolicies, updateUnderstandingSourceRun, upsertUnderstandingSourceGrant } from "../user-context/sources/repository.js";
 import { allowsRemoteSourceProcessing } from "../user-context/sources/processing-policy.js";

package/dist/src/workflows/service/workflow-run-service.js

@@ -4,8 +4,8 @@ import { init_conversation_repository, resolveRoutedConversation } from "../../s
 import { getDefaultAgentId, init_resolve_route } from "../../routing/resolve-route.js";
 import { extractProfileAgentId, resolveEffectiveAgentProfile, resolveEffectiveAgentProfileForSession } from "../../config/agent-profile.js";
 import { init_providers, resolveModel } from "../../providers/index.js";
-import { TaskConversationRepository } from "../../tasks/task-conversation-repository.js";
 import { getProjectWorkspacePathForSession } from "../../projects/workspace.js";
+import { TaskConversationRepository } from "../../tasks/task-conversation-repository.js";
 import { automationEventPublisher, getAutomationEventForRun } from "../../automations/events/event-repository.js";
 import { isTerminalWorkflowRunStatus } from "../domain/run.js";
 import { validateWorkflowJsonSchema } from "../domain/schema-validation.js";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xopcai/xopc",
-  "version": "0.0.364",
+  "version": "0.0.366",
   "description": "Local-first personal AI for the one-person company, with durable context, tasks, workflows, automations, and local or cloud models.",
   "type": "module",
   "main": "dist/src/index.js",
@@ -162,11 +162,11 @@
     "vitest": "4.1.5",
     "vue": "^3.5.35",
     "@xopcai/browser-control-contract": "0.1.0",
-    "@xopcai/endpoint-tools-client": "0.0.0",
     "@xopcai/endpoint-tools-protocol": "0.0.0",
+    "@xopcai/endpoint-tools-client": "0.0.0",
+    "@xopcai/gateway-contract": "0.0.0",
     "@xopcai/computer-control-contract": "0.1.0",
     "@xopcai/realtime-client": "0.0.0",
-    "@xopcai/gateway-contract": "0.0.0",
     "@xopcai/realtime-protocol": "0.0.0"
   },
   "scripts": {

package/dist/gateway/static/root/assets/gateway-realtime-store-DBvCRclh.js

@@ -1 +0,0 @@
-import{p as e}from"./locale-store-DId6-v2R.js";var t=e(()=>({connectionState:`idle`,error:null}));function n(e){t.setState(e)}export{t as n,n as t};