@xopcai/xopc 0.0.362 → 0.0.364

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

@@ -70,7 +70,12 @@ var DurableVoiceAgentBroker = class {
 		return {
 			taskId: row.id,
 			runId,
-			events: this.events(row.id, runId, afterSeq, input.signal)
+			events: this.events(row.id, runId, afterSeq, input.signal, [
+				"completed",
+				"cancelled",
+				"failed",
+				"interrupted"
+			].includes(row.status))
 		};
 	}
 	async cancel(taskId) {
@@ -80,7 +85,7 @@ var DurableVoiceAgentBroker = class {
 		this.runByTask.delete(taskId);
 		return true;
 	}
-	async *events(taskId, runId, afterSeq, signal) {
+	async *events(taskId, runId, afterSeq, signal, ended) {
 		const queued = [];
 		let wake;
 		const push = (event) => {
@@ -96,6 +101,10 @@ var DurableVoiceAgentBroker = class {
 		};
 		signal.addEventListener("abort", abort, { once: true });
 		try {
+			if (ended && !subscription.initial.some((delivery) => {
+				const event = eventData(delivery);
+				return event?.type === "run_end" || event?.type === "stream_end";
+			})) throw new Error("Completed voice task event history is unavailable");
 			while (!signal.aborted) {
 				if (!queued.length) await new Promise((resolve) => {
 					wake = resolve;

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

@@ -3,7 +3,7 @@ import { init_logger } from "../../utils/logger.js";
 import { speakStream } from "../tts/speak-core.js";
 import { AudioPlaybackWindow } from "./audio-playback-window.js";
 import { TurnCoordinator } from "./turnPolicy.js";
-import { isLikelyPlaybackEcho } from "./playback-echo.js";
+import { PlaybackEchoCandidates } from "./playback-echo.js";
 import { PcmFrameBuffer } from "./pcmFrameBuffer.js";
 import { SpeakableSegmenter } from "./speakable-segmenter.js";
 import crypto from "node:crypto";
@@ -36,6 +36,7 @@ function createAgentVoiceEngine(options) {
 	let finalCount = 0;
 	let committing = false;
 	let recentPlayback;
+	const echoCandidates = new PlaybackEchoCandidates();
 	const finalizedUtterances = /* @__PURE__ */ new Set();
 	let pendingTurn;
 	let taskUpdateQueued = false;
@@ -89,6 +90,11 @@ function createAgentVoiceEngine(options) {
 	function cancelActiveResponse(reason) {
 		const response = activeResponse;
 		if (!response) return false;
+		if (response.audioStarted && response.audibleText) recentPlayback = {
+			responseId: response.id,
+			text: response.audibleText,
+			expiresAt: Date.now() + PLAYBACK_ECHO_TAIL_MS
+		};
 		activeResponse = void 0;
 		response.abortController.abort(reason);
 		send("response.cancelled", {
@@ -267,13 +273,11 @@ function createAgentVoiceEngine(options) {
 				conversationId: claim.request.conversationId,
 				expectedTranscriptId: claim.conversationSessionId,
 				signal: response.abortController.signal,
-				...updateClientMessageId ? {
-					clientMessageId: updateClientMessageId,
-					origin: {
-						type: "system",
-						source: "task_update"
-					}
-				} : {}
+				clientMessageId: updateClientMessageId ?? `voice:${claim.conversationSessionId}:${claim.sessionId}:${inputGeneration}:${turnId}`,
+				...updateClientMessageId ? { origin: {
+					type: "system",
+					source: "task_update"
+				} } : {}
 			});
 			response.taskId = task.taskId;
 			send("task.created", {
@@ -438,6 +442,7 @@ function createAgentVoiceEngine(options) {
 		if (event.type === "ready" || event.type === "usage") return;
 		if (event.type === "error") {
 			turn.reset();
+			echoCandidates.clear();
 			log.warn({
 				err: event.error,
 				sessionId: claim.sessionId,
@@ -454,6 +459,7 @@ function createAgentVoiceEngine(options) {
 		if (muted || finalizedUtterances.has(event.utteranceId)) return;
 		if (event.type === "speech_started") {
 			lastUserSpeechAt = Date.now();
+			echoCandidates.remember(event.utteranceId, activeResponse?.audioStarted ? activeResponse.audibleText : recentPlayback && recentPlayback.expiresAt >= Date.now() ? recentPlayback.text : void 0);
 			if (claim.request.purpose === "conversation") {
 				if (pendingTurn) {
 					pendingTurn.cancelled = true;
@@ -490,11 +496,11 @@ function createAgentVoiceEngine(options) {
 				responseId: activeResponse.id,
 				text: activeResponse.audibleText
 			} : recentPlayback && recentPlayback.expiresAt >= Date.now() ? recentPlayback : void 0;
-			if (claim.request.purpose === "conversation" && playback && isLikelyPlaybackEcho(text, playback.text)) {
+			if (claim.request.purpose === "conversation" && echoCandidates.matches(event.utteranceId, text, playback?.text)) {
 				bufferFinal(event.utteranceId, "");
 				log.debug({
 					sessionId: claim.sessionId,
-					responseId: playback.responseId,
+					responseId: playback?.responseId,
 					transcriptCharacters: text.length
 				}, "Ignored finalized transcription matching recent voice playback");
 				return;
@@ -517,6 +523,7 @@ function createAgentVoiceEngine(options) {
 	}
 	function discardInput() {
 		turn.reset();
+		echoCandidates.clear();
 		pendingTurn = void 0;
 		inputGeneration += 1;
 		queuedTurns = 0;
@@ -623,6 +630,7 @@ function createAgentVoiceEngine(options) {
 			if (closing) return closing;
 			closed = true;
 			turn.reset();
+			echoCandidates.clear();
 			cancelActiveResponse("session_closed");
 			sttAbort.abort("session_closed");
 			sttSession?.abort("session_closed");

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

@@ -2,7 +2,7 @@ import { createLogger } from "../../utils/logger/index.js";
 import { init_logger } from "../../utils/logger.js";
 import { AudioPlaybackWindow } from "./audio-playback-window.js";
 import { TurnCoordinator } from "./turnPolicy.js";
-import { isLikelyPlaybackEcho } from "./playback-echo.js";
+import { PlaybackEchoCandidates } from "./playback-echo.js";
 import { PcmFrameBuffer } from "./pcmFrameBuffer.js";
 import crypto from "node:crypto";
 import WebSocket from "ws";
@@ -45,6 +45,7 @@ function createOmniVoiceEngine(options) {
 	let failed = false;
 	let platformRequestId;
 	let active;
+	const echoCandidates = new PlaybackEchoCandidates();
 	let recentPlayback;
 	let waitingForCancellation;
 	let responseCreateQueued = false;
@@ -147,6 +148,10 @@ function createOmniVoiceEngine(options) {
 	const cancel = (reason) => {
 		const response = active;
 		if (!response) return false;
+		if (response.audio && response.text) recentPlayback = {
+			text: response.text,
+			expiresAt: Date.now() + PLAYBACK_ECHO_TAIL_MS
+		};
 		active = void 0;
 		response.abort.abort(reason);
 		response.release();
@@ -234,6 +239,7 @@ function createOmniVoiceEngine(options) {
 		failed = true;
 		clearTimeout(responseStartTimer);
 		turn.reset();
+		echoCandidates.clear();
 		responseCreateQueued = false;
 		log.warn({
 			sessionId: options.callId,
@@ -259,6 +265,7 @@ function createOmniVoiceEngine(options) {
 	function discardInput() {
 		clearTimeout(responseStartTimer);
 		turn.reset();
+		echoCandidates.clear();
 		responseCreateQueued = false;
 		speaking.clear();
 		turnSettled = false;
@@ -358,6 +365,7 @@ function createOmniVoiceEngine(options) {
 								return;
 							}
 							if (recorded.has(String(event.item_id)) || pendingInputs.has(String(event.item_id))) return;
+							echoCandidates.remember(String(event.item_id), active?.audio ? active.text : recentPlayback && recentPlayback.expiresAt >= Date.now() ? recentPlayback.text : void 0);
 							inputBlocked = false;
 							pendingInputs.add(String(event.item_id));
 							clearTimeout(responseStartTimer);
@@ -381,7 +389,7 @@ function createOmniVoiceEngine(options) {
 							speaking.delete(event.item_id);
 							const text = event.transcript.trim();
 							const playbackText = active?.audio ? active.text : recentPlayback && recentPlayback.expiresAt >= Date.now() ? recentPlayback.text : void 0;
-							if (playbackText && isLikelyPlaybackEcho(text, playbackText)) {
+							if (echoCandidates.matches(event.item_id, text, playbackText)) {
 								turn.final(event.item_id, "");
 								log.debug({
 									sessionId: options.callId,
@@ -582,6 +590,7 @@ function createOmniVoiceEngine(options) {
 			clearTimeout(uploadTimer);
 			clearTimeout(responseStartTimer);
 			turn.reset();
+			echoCandidates.clear();
 			speaking.clear();
 			inputQueue = [];
 			queuedInputBytes = 0;

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

@@ -1,6 +1,6 @@
 import { init_agent_scope, resolveAgentProfileDir } from "../../agent/agent-scope.js";
-import { resolveEffectiveAgentProfileForSession } from "../../config/agent-profile.js";
 import { DEFAULT_SOUL_FILENAME } from "../../agent/context/workspace.js";
+import { resolveEffectiveAgentProfileForSession } from "../../config/agent-profile.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/playback-echo.d.ts

@@ -1,2 +1,9 @@
 /** Conservatively identifies finalized ASR that is already in audible assistant output. */
 export declare function isLikelyPlaybackEcho(transcript: string, assistantText: string): boolean;
+/** Retains the playback at speech onset while a provider finishes delayed ASR. */
+export declare class PlaybackEchoCandidates {
+    private readonly candidates;
+    remember(utteranceId: string, text: string | undefined): void;
+    matches(utteranceId: string, transcript: string, currentPlayback?: string): boolean;
+    clear(): void;
+}

package/dist/src/voice/realtime/playback-echo.js

@@ -28,5 +28,27 @@ function isLikelyPlaybackEcho(transcript, assistantText) {
 	if (heard.length < (containsCjk ? 6 : 10)) return false;
 	return closestSubstringEditDistance(heard, spoken) / Array.from(heard).length <= .22;
 }
+/** Retains the playback at speech onset while a provider finishes delayed ASR. */
+var PlaybackEchoCandidates = class {
+	candidates = /* @__PURE__ */ new Map();
+	remember(utteranceId, text) {
+		if (!text || this.candidates.has(utteranceId)) return;
+		const now = Date.now();
+		for (const [id, candidate] of this.candidates) if (candidate.expiresAt < now) this.candidates.delete(id);
+		if (this.candidates.size >= 256) this.candidates.delete(this.candidates.keys().next().value);
+		this.candidates.set(utteranceId, {
+			text: text.slice(-8e3),
+			expiresAt: now + 3e4
+		});
+	}
+	matches(utteranceId, transcript, currentPlayback) {
+		const candidate = this.candidates.get(utteranceId);
+		this.candidates.delete(utteranceId);
+		return Boolean(currentPlayback && isLikelyPlaybackEcho(transcript, currentPlayback) || candidate && candidate.expiresAt >= Date.now() && isLikelyPlaybackEcho(transcript, candidate.text));
+	}
+	clear() {
+		this.candidates.clear();
+	}
+};
 //#endregion
-export { isLikelyPlaybackEcho };
+export { PlaybackEchoCandidates, isLikelyPlaybackEcho };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xopcai/xopc",
-  "version": "0.0.362",
+  "version": "0.0.364",
   "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",
@@ -75,6 +75,9 @@
     "@inquirer/prompts": "^8.5.0",
     "@modelcontextprotocol/sdk": "1.32.1",
     "@mozilla/readability": "^0.6.0",
+    "@opentelemetry/api": "^1.9.1",
+    "@opentelemetry/context-async-hooks": "^2.12.0",
+    "@opentelemetry/sdk-trace-base": "^2.12.0",
     "@sinclair/typebox": "^0.34.49",
     "@trycua/cua-driver": "0.28.2",
     "@vscode/ripgrep": "^1.18.0",
@@ -158,12 +161,12 @@
     "vitepress": "^1.6.4",
     "vitest": "4.1.5",
     "vue": "^3.5.35",
-    "@xopcai/endpoint-tools-client": "0.0.0",
     "@xopcai/browser-control-contract": "0.1.0",
-    "@xopcai/computer-control-contract": "0.1.0",
+    "@xopcai/endpoint-tools-client": "0.0.0",
     "@xopcai/endpoint-tools-protocol": "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-DS4WzQyF.js

@@ -1 +0,0 @@
-import{p as e}from"./locale-store-BbwWB0ra.js";var t=e(()=>({connectionState:`idle`,error:null}));function n(e){t.setState(e)}export{t as n,n as t};