@xopcai/xopc 0.0.229 → 0.0.230

package/dist/src/agent/inbound/turn-dispatcher.js

@@ -1,8 +1,8 @@
 import { queueEmbeddedSteer } from "../embedded/runs.js";
 import { buildDirectUserMessageContent } from "../service/build-direct-message-content.js";
+import { maybeEmitWebchatTts } from "../service/webchat-tts.js";
 import { runProcessDirectStreaming } from "../service/process-direct-streaming.js";
 import { runProcessDirect } from "../service/process-direct-one-shot.js";
-import { maybeEmitWebchatTts } from "../service/webchat-tts.js";
 //#region src/agent/inbound/turn-dispatcher.ts
 var TurnDispatcher = class {
 	cfg;

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

@@ -13,6 +13,7 @@ import { isSessionSourceBinding, summarizeSourceContext } from "../source-contex
 import { injectSourceContextsIntoUserMessage } from "../source-context/injector.js";
 import { AsyncQueue, AsyncQueueOverflowError } from "./async-queue.js";
 import { hydratePerTurnState, runDirectAgentTurn, tryRunSlashCommand } from "./direct-turn-helpers.js";
+import { isSuccessfulWebchatTtsToolEvent } from "./webchat-tts.js";
 //#region src/agent/service/process-direct-streaming.ts
 const MAX_BUFFERED_STREAM_EVENTS = 512;
 const LOSSY_STREAM_EVENT_TYPES = new Set([
@@ -151,6 +152,7 @@ async function* runProcessDirectStreaming(deps, input) {
 	let userAborted = false;
 	let abortHandled = false;
 	let inboundVoice = false;
+	let explicitTtsEmitted = false;
 	let ranSlashCommand = false;
 	let mergedUserText = input.content;
 	let webchatSlashReceipt;
@@ -299,6 +301,7 @@ async function* runProcessDirectStreaming(deps, input) {
 					runId: input.runId,
 					onEvent: (embeddedEvent) => {
 						const event = { ...embeddedEvent };
+						if (isSuccessfulWebchatTtsToolEvent(event)) explicitTtsEmitted = true;
 						if (event.type === "error" && typeof event.content === "string") event.content = formatStreamError(event.content);
 						pushVisible(event);
 					}
@@ -329,7 +332,7 @@ async function* runProcessDirectStreaming(deps, input) {
 				sessionKey,
 				droppedEvents: queue.droppedCount
 			}, "Coalesced stream progress events because the consumer fell behind");
-			if (!userAborted && !streamOverflowed && channel === "webchat") try {
+			if (!userAborted && !streamOverflowed && channel === "webchat" && !explicitTtsEmitted) try {
 				const ttsAudioEvent = await deps.maybeEmitWebchatTts(sessionKey, inboundVoice);
 				if (ttsAudioEvent) queue.push(ttsAudioEvent);
 			} catch (ttsErr) {

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

@@ -15,6 +15,8 @@ export type WebchatTtsDeps = {
         warn: (obj: Record<string, unknown>, msg: string) => void;
     };
 };
+/** True when the explicit TTS tool already produced durable Webchat audio. */
+export declare function isSuccessfulWebchatTtsToolEvent(event: unknown): boolean;
 /**
  * Generate TTS for webchat when config allows; persist under `{stateDir}/media/tts/`.
  */

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

@@ -3,9 +3,22 @@ import { speak } from "../../voice/tts/speak-core.js";
 import { mergeTtsConfigFromAppConfig } from "../../voice/tts/merge-config.js";
 import { getChannelOutputFormat, shouldUseTTS } from "../../voice/tts/service.js";
 import { compressAudio } from "../../voice/tts/audio.js";
-import "../../voice/tts/index.js";
 import { persistOutboundTtsAudio } from "../../channels/attachments/outbound-tts-persist.js";
+import "../../voice/tts/index.js";
 //#region src/agent/service/webchat-tts.ts
+/** True when the explicit TTS tool already produced durable Webchat audio. */
+function isSuccessfulWebchatTtsToolEvent(event) {
+	if (!event || typeof event !== "object" || Array.isArray(event)) return false;
+	const row = event;
+	if (row.type !== "tool_execution_end" || row.toolName !== "text_to_speech" || row.isError === true) return false;
+	const result = row.result && typeof row.result === "object" && !Array.isArray(row.result) ? row.result : null;
+	const details = result?.details && typeof result.details === "object" && !Array.isArray(result.details) ? result.details : null;
+	return Array.isArray(details?.media) && details.media.some((item) => {
+		if (!item || typeof item !== "object" || Array.isArray(item)) return false;
+		const media = item;
+		return typeof media.uri === "string" && media.uri.trim().length > 0;
+	});
+}
 /**
 * Generate TTS for webchat when config allows; persist under `{stateDir}/media/tts/`.
 */
@@ -57,4 +70,4 @@ async function appendMediaToLastAssistant(sessionStore, sessionKey, ref) {
 	}
 }
 //#endregion
-export { appendMediaToLastAssistant, maybeEmitWebchatTts };
+export { appendMediaToLastAssistant, isSuccessfulWebchatTtsToolEvent, maybeEmitWebchatTts };

package/dist/src/agent/tools/tts-tool.js

@@ -4,6 +4,7 @@ import { speak } from "../../voice/tts/speak-core.js";
 import { mergeTtsConfigFromAppConfig } from "../../voice/tts/merge-config.js";
 import { getChannelOutputFormat } from "../../voice/tts/service.js";
 import { compressAudio } from "../../voice/tts/audio.js";
+import { persistOutboundTtsAudio } from "../../channels/attachments/outbound-tts-persist.js";
 import { Type } from "@sinclair/typebox";
 //#region src/agent/tools/tts-tool.ts
 init_logger();
@@ -72,30 +73,35 @@ function createTextToSpeechTool(deps) {
 				});
 				const wavTarget = outFmt.format === "mp3" ? "mp3" : "opus";
 				const { buffer: compressedAudio, format: compressedFormat } = await compressAudio(Buffer.from(result.audio), result.format, wavTarget);
-				const dataUrl = `data:${compressedFormat === "opus" ? "audio/ogg" : compressedFormat === "mp3" || compressedFormat === "mpeg" ? "audio/mpeg" : `audio/${compressedFormat}`};base64,${compressedAudio.toString("base64")}`;
-				const msg = {
-					channel: ctx.channel,
-					chat_id: ctx.chatId,
-					content: "",
-					mediaUrl: dataUrl,
-					mediaType: "audio",
-					audioAsVoice: outFmt.voiceCompatible
-				};
-				await deps.bus.publishOutbound(msg);
+				const mimeType = compressedFormat === "opus" ? "audio/ogg" : compressedFormat === "mp3" || compressedFormat === "mpeg" ? "audio/mpeg" : `audio/${compressedFormat}`;
+				const media = ctx.channel === "webchat" ? await persistOutboundTtsAudio(compressedAudio, compressedFormat) : void 0;
+				if (!media) {
+					const msg = {
+						channel: ctx.channel,
+						chat_id: ctx.chatId,
+						content: "",
+						mediaUrl: `data:${mimeType};base64,${compressedAudio.toString("base64")}`,
+						mediaType: "audio",
+						audioAsVoice: outFmt.voiceCompatible
+					};
+					await deps.bus.publishOutbound(msg);
+				}
 				log.info({
 					provider: result.provider,
 					format: compressedFormat,
+					channel: ctx.channel,
 					textLength: params.text.length,
 					audioSize: compressedAudio.length
-				}, "TTS tool sent audio");
+				}, ctx.channel === "webchat" ? "TTS tool attached audio" : "TTS tool sent audio");
 				return {
 					content: [{
 						type: "text",
-						text: `✅ Sent voice message (${result.provider}, ${compressedFormat}).`
+						text: ctx.channel === "webchat" ? `✅ Attached voice message (${result.provider}, ${compressedFormat}).` : `✅ Sent voice message (${result.provider}, ${compressedFormat}).`
 					}],
 					details: {
 						provider: result.provider,
-						format: compressedFormat
+						format: compressedFormat,
+						...media ? { media: [media] } : {}
 					}
 				};
 			} catch (error) {

package/dist/src/gateway/chat-stream/mapper.js

@@ -342,6 +342,17 @@ var ChatStreamMapper = class {
 			errorMessage: event.isError ? result?.text ?? extractText(result?.content) : void 0
 		})];
 		const details = asRecord(fullResult?.details);
+		if (toolName === "text_to_speech" && !event.isError && details) {
+			const audio = (Array.isArray(details.media) ? details.media : []).map(asRecord).find((item) => item?.type === "voice" || String(item?.mimeType ?? "").startsWith("audio/"));
+			const uri = typeof audio?.uri === "string" ? audio.uri.trim() : "";
+			if (uri) events.push(this.make("tts_audio", {
+				uri,
+				mimeType: typeof audio?.mimeType === "string" ? audio.mimeType : "audio/mpeg",
+				name: typeof audio?.name === "string" ? audio.name : "voice.mp3",
+				attachTo: "last_assistant",
+				messageId
+			}));
+		}
 		if (toolName === "exec_command" && details) events.push(this.make("command_completed", {
 			messageId,
 			toolCallId,

package/dist/src/gateway/hono/routes/voice.js

@@ -14,8 +14,8 @@ import "../../../voice/stt/index.js";
 import { listSttProvidersForApi } from "../../../voice/stt/list-providers.js";
 import "../../../voice/tts/index.js";
 import { createGatewayRouteLogger } from "../lib/route-logger.js";
-import { listLocalVoiceModelStatuses, removeLocalVoiceModel, startLocalVoiceModelInstall } from "../../../voice/local/model-manager.js";
 import { listTtsProvidersForApi } from "../../../voice/tts/list-providers.js";
+import { listLocalVoiceModelStatuses, removeLocalVoiceModel, startLocalVoiceModelInstall } from "../../../voice/local/model-manager.js";
 //#region src/gateway/hono/routes/voice.ts
 init_providers();
 const log = createGatewayRouteLogger("Voice");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xopcai/xopc",
-  "version": "0.0.229",
+  "version": "0.0.230",
   "description": "Local-first personal AI assistant with reviewable user understanding, durable tasks, workflows, automations, and local or cloud models.",
   "type": "module",
   "main": "dist/src/index.js",

package/dist/gateway/static/root/assets/settings-advanced-gate-JFCbnCL2.js

@@ -1 +0,0 @@
-import{Or as e}from"./index-BNWhgmjs.js";function t(t){let{children:n,fallback:r}=t,i=r===void 0?null:r;return e()?n:i}export{t};