@xopcai/xopc 0.0.241 → 0.0.243

package/dist/src/config/schema.js

@@ -8,6 +8,7 @@ import { init_public_url, validatePublicUrl } from "./public-url.js";
 import { TelegramAccountConfigSchema, TelegramConfigSchema, TelegramGroupConfigSchema, TelegramTopicConfigSchema, init_config_schema } from "../../extensions/telegram/src/config-schema.js";
 import { WeixinAccountConfigSchema, WeixinConfigSchema, init_config_schema as init_config_schema$1 } from "../../extensions/weixin/src/config-schema.js";
 import { z } from "zod";
+import semver from "semver";
 //#region src/config/schema.ts
 function getAgentDefaultModelRef(config) {
 	return config.agents.list.find((entry) => entry.enabled !== false && entry.id === config.agents.default)?.models?.chat?.primary ?? config.agents.defaults.models.chat.primary;
@@ -36,7 +37,7 @@ function parseModelRef(ref) {
 		model: trimmed.slice(idx + 1).trim()
 	};
 }
-var AgentModelRefSchema, AgentImageGenerationModelSchema, AgentModelsSchema, AgentsConfigSchema, BrowserCloudConfigSchema, BrowserExtensionConfigSchema, BrowserCloakConfigSchema, BrowserConfigSchema, BindingMatchSchema, BindingRuleSchema, BindingsConfigSchema, SessionDmScopeSchema, SessionStorageConfigSchema, SessionResetModeSchema, SessionResetConfigSchema, SessionResetByTypeSchema, SessionScopeSchema, SessionConfigSchema, ChannelsConfigSchema, SearchProviderEntrySchema, WebSearchConfigSchema, WebsiteBlocklistSchema, WebToolsConfigSchema, ToolsConfigSchema, RuntimePreferenceSchema, RuntimeProvisionSchema, LanguageRuntimeSchema, DEFAULT_RUNTIME_TOOLS_CONFIG, RuntimeToolsConfigSchema, GatewayAuthRateLimitSchema, GatewaySecuritySchema, GatewayTrustedProxySchema, GatewayAuthSchema, GatewayTailscaleConsentSchema, GatewayTailscaleSchema, GatewayRemoteSchema, GatewayTlsSchema, GatewayModeSchema, HeartbeatConfigSchema, GatewayChannelConnectDeferModeSchema, TunnelConsentSchema, TunnelConfigSchema, WorkspaceImportConfigSchema, WorkspaceConfigSchema, GatewayBindModeSchema, GatewayWebchatConfigSchema, GatewayConfigSchema, ModelsDevConfigSchema, ModelCatalogConfigSchema, SttProviderConfigCatchallSchema, STTProviderConfigSchema, STTFallbackConfigSchema, MediaUnderstandingCapabilitiesSchema, MediaUnderstandingModelSchema, STTConfigSchema, TtsProviderConfigCatchallSchema, TTSProviderConfigSchema, TTSFallbackConfigSchema, TTSModelOverridesConfigSchema, TTSEdgeConfigSchema, TTSSummarizationConfigSchema, TTSConfigSchema, VoiceRefinementConfigSchema, VoiceConfigSchema, MessagesConfigSchema, ToolsMediaAudioConfigSchema, ToolsMediaConfigSchema, ProviderAzureConfigSchema, ProviderRequestOverridesSchema, ProviderAuthConfigSchema, ProvidersConfigSchema, ExtensionSecurityConfigSchema, ExtensionSlotsConfigSchema, ExtensionsConfigSchema, UpdateAutoConfigSchema, UpdateConfigSchema, CommandsConfigSchema, TuiConfigSchema, McpHttpUrlSchema, ConnectorSecretReferenceSchema, McpConfigScalarSchema, McpOAuthConfigSchema, McpServerSchema, McpConfigSchema, ConnectorsConfigSchema, ExperimentalConfigSchema, ConfigSchema;
+var AgentModelRefSchema, AgentImageGenerationModelSchema, AgentModelsSchema, AgentsConfigSchema, BrowserCloudConfigSchema, BrowserExtensionConfigSchema, BrowserCloakConfigSchema, BrowserConfigSchema, BindingMatchSchema, BindingRuleSchema, BindingsConfigSchema, SessionDmScopeSchema, SessionStorageConfigSchema, SessionResetModeSchema, SessionResetConfigSchema, SessionResetByTypeSchema, SessionScopeSchema, SessionConfigSchema, ChannelsConfigSchema, SearchProviderEntrySchema, WebSearchConfigSchema, WebsiteBlocklistSchema, WebToolsConfigSchema, ToolsConfigSchema, RuntimePreferenceSchema, RuntimeProvisionSchema, RuntimeVersionRequestSchema, LanguageRuntimeSchema, DEFAULT_RUNTIME_TOOLS_CONFIG, RuntimeToolsConfigSchema, GatewayAuthRateLimitSchema, GatewaySecuritySchema, GatewayTrustedProxySchema, GatewayAuthSchema, GatewayTailscaleConsentSchema, GatewayTailscaleSchema, GatewayRemoteSchema, GatewayTlsSchema, GatewayModeSchema, HeartbeatConfigSchema, GatewayChannelConnectDeferModeSchema, TunnelConsentSchema, TunnelConfigSchema, WorkspaceImportConfigSchema, WorkspaceConfigSchema, GatewayBindModeSchema, GatewayWebchatConfigSchema, GatewayConfigSchema, ModelsDevConfigSchema, ModelCatalogConfigSchema, SttProviderConfigCatchallSchema, STTProviderConfigSchema, STTFallbackConfigSchema, MediaUnderstandingCapabilitiesSchema, MediaUnderstandingModelSchema, STTConfigSchema, TtsProviderConfigCatchallSchema, TTSProviderConfigSchema, TTSFallbackConfigSchema, TTSModelOverridesConfigSchema, TTSEdgeConfigSchema, TTSSummarizationConfigSchema, TTSConfigSchema, VoiceRefinementConfigSchema, VoiceConfigSchema, MessagesConfigSchema, ToolsMediaAudioConfigSchema, ToolsMediaConfigSchema, ProviderAzureConfigSchema, ProviderRequestOverridesSchema, ProviderAuthConfigSchema, ProvidersConfigSchema, ExtensionSecurityConfigSchema, ExtensionSlotsConfigSchema, ExtensionsConfigSchema, UpdateAutoConfigSchema, UpdateConfigSchema, CommandsConfigSchema, TuiConfigSchema, McpHttpUrlSchema, ConnectorSecretReferenceSchema, McpConfigScalarSchema, McpOAuthConfigSchema, McpServerSchema, McpConfigSchema, ConnectorsConfigSchema, ExperimentalConfigSchema, ConfigSchema;
 var init_schema = __esmMin((() => {
 	init_agent_config();
 	init_cache_dir_policy();
@@ -240,9 +241,10 @@ var init_schema = __esmMin((() => {
 		"on-demand",
 		"disabled"
 	]);
+	RuntimeVersionRequestSchema = z.string().trim().min(1).max(128).refine((value) => !value.includes("/") && !value.includes("\\") && !value.includes("\0"), "Runtime version must not contain path separators").refine((value) => semver.validRange(value) !== null, "Runtime version must be a valid SemVer version or range");
 	LanguageRuntimeSchema = z.object({
 		enabled: z.boolean().default(true),
-		version: z.string().min(1).optional(),
+		version: RuntimeVersionRequestSchema.optional(),
 		preference: RuntimePreferenceSchema.default("managed-first"),
 		provision: RuntimeProvisionSchema.default("on-demand")
 	});
@@ -280,7 +282,7 @@ var init_schema = __esmMin((() => {
 		}),
 		uv: z.object({
 			enabled: z.boolean().default(true),
-			version: z.string().min(1).optional()
+			version: RuntimeVersionRequestSchema.optional()
 		}).default({ enabled: true }),
 		download: z.object({
 			bundleDir: z.string().min(1).optional(),
@@ -798,7 +800,37 @@ activityDetailDefault: z.enum([
 	VoiceConfigSchema = z.object({
 		languageMode: z.enum(["auto", "manual"]).optional(),
 		language: z.enum(["en", "zh"]).optional(),
-		input: z.object({ refinement: VoiceRefinementConfigSchema.optional() }).strict().optional()
+		input: z.object({ refinement: VoiceRefinementConfigSchema.optional() }).strict().optional(),
+		realtime: z.object({
+			enabled: z.boolean().default(false),
+			defaultEngine: z.enum(["agent", "omni"]).default("agent"),
+			silenceDurationMs: z.number().int().min(300).max(2e3).default(700),
+			idleTimeoutMs: z.number().int().min(1e4).max(3e5).default(6e4),
+			maxDictationMs: z.number().int().min(1e4).max(36e5).default(6e5),
+			maxConversationMs: z.number().int().min(1e4).max(72e5).default(36e5),
+			bargeIn: z.boolean().default(true),
+			/** Conversation output selection; credentials stay in the provider configuration. */
+			tts: z.object({
+				provider: z.enum(["alibaba", "xopc-cloud"]),
+				voice: z.string().trim().min(1).max(200).optional()
+			}).strict().optional(),
+			omni: z.object({
+				provider: z.enum(["alibaba", "xopc-cloud"]),
+				model: z.string().min(1).max(200),
+				voice: z.string().min(1).max(100),
+				baseUrl: z.string().url().optional(),
+				apiKey: z.string().min(1).optional(),
+				instructions: z.string().max(8e3).default("You are a friendly voice companion. Keep replies conversational and concise. You cannot execute tools or perform actions.")
+			}).strict().optional()
+		}).strict().default({
+			enabled: false,
+			defaultEngine: "agent",
+			silenceDurationMs: 700,
+			idleTimeoutMs: 6e4,
+			maxDictationMs: 6e5,
+			maxConversationMs: 36e5,
+			bargeIn: true
+		})
 	}).strict().optional();
 	MessagesConfigSchema = z.object({ 
 	/** Voice (text-to-speech) output configuration. */

package/dist/src/extensions/install.js

@@ -8,8 +8,8 @@ import { collectExtensionPackageDependencyIssues } from "./package-contract.js";
 import { dirname, isAbsolute, join, resolve, sep } from "path";
 import { cpSync, existsSync, mkdirSync, mkdtempSync, readFileSync, readdirSync, rmSync, writeFileSync } from "fs";
 import { execFileSync } from "node:child_process";
-import AdmZip from "adm-zip";
 import semver from "semver";
+import AdmZip from "adm-zip";
 import { tmpdir } from "os";
 //#region src/extensions/install.ts
 /**

package/dist/src/files/file-service.d.ts

@@ -40,5 +40,6 @@ export declare class FileSpaceService {
         resource: FileResource;
         absolutePath: string;
     }>;
+    rootResource(spaceId: string): Promise<FileResource>;
     children(spaceId: string, path?: string): Promise<FileResource[]>;
 }

package/dist/src/files/file-service.js

@@ -289,6 +289,10 @@ var FileSpaceService = class {
 			absolutePath
 		};
 	}
+	async rootResource(spaceId) {
+		const space = await this.get(spaceId);
+		return fileResourceFromPath(space, await resolveFilePath(space.root, ""));
+	}
 	async children(spaceId, path = "") {
 		const space = await this.get(spaceId);
 		const directory = await resolveFilePath(space.root, path);

package/dist/src/gateway/chat-stream/thinking-delta-coalescer.js

@@ -35,7 +35,12 @@ async function* coalesceThinkingDeltas(source, options = {}) {
 	if (!Number.isFinite(windowMs) || windowMs <= 0) throw new Error("thinking delta coalescer windowMs must be greater than zero");
 	if (!Number.isInteger(maxChars) || maxChars <= 0) throw new Error("thinking delta coalescer maxChars must be a positive integer");
 	const iterator = source[Symbol.asyncIterator]();
-	let next = iterator.next();
+	const readNext = () => {
+		const result = iterator.next();
+		result.catch(() => {});
+		return result;
+	};
+	let next = readNext();
 	let pending;
 	let deadlineAt = 0;
 	try {
@@ -43,7 +48,7 @@ async function* coalesceThinkingDeltas(source, options = {}) {
 			if (!pending) {
 				const result = await next;
 				if (result.done) break;
-				next = iterator.next();
+				next = readNext();
 				if (!isThinkingDelta(result.value)) {
 					yield result.value;
 					continue;
@@ -68,7 +73,7 @@ async function* coalesceThinkingDeltas(source, options = {}) {
 				pending = void 0;
 				break;
 			}
-			next = iterator.next();
+			next = readNext();
 			const event = result.value;
 			if (!isThinkingDelta(event) || !belongsToSameThinkingStream(pending, event)) {
 				yield pending;

package/dist/src/gateway/hono/lib/config-payload.d.ts

@@ -274,17 +274,7 @@ export declare function buildSafeWebConfigPayload(service: GatewayService, optio
     };
     stt: unknown;
     tts: unknown;
-    voice: {
-        languageMode?: "manual" | "auto";
-        language?: "en" | "zh";
-        input?: {
-            refinement?: {
-                mode: "off" | "custom" | "punctuation" | "light";
-                model?: string;
-                customInstruction?: string;
-            };
-        };
-    };
+    voice: unknown;
     tools: {
         web: {
             region: "cn" | "global" | null;

package/dist/src/gateway/hono/lib/config-payload.js

@@ -13,7 +13,7 @@ import { GENERIC_MASKED_SECRET, maskSecretLength } from "./mask-secret-length.js
 import { safeToolsWebForGet } from "../../config-tools-web.js";
 import { maskTunnelSecretForWeb } from "../../../tunnel/env.js";
 import { buildSafeProvidersConfigForWeb } from "./safe-providers-config.js";
-import { maskSttConfigForWeb, maskTtsConfigForWeb } from "./safe-voice-config.js";
+import { maskRealtimeVoiceConfigForWeb, maskSttConfigForWeb, maskTtsConfigForWeb } from "./safe-voice-config.js";
 //#region src/gateway/hono/lib/config-payload.ts
 init_agent_scope();
 init_channel_catalog_service();
@@ -195,7 +195,7 @@ async function buildSafeWebConfigPayload(service, options = {}) {
 		update: { ...resolveUpdateConfigForWeb(config) },
 		stt: maskSttConfigForWeb(config.tools?.media?.audio),
 		tts: maskTtsConfigForWeb(config.messages?.tts),
-		voice: config.voice,
+		voice: maskRealtimeVoiceConfigForWeb(config.voice),
 		tools: safeToolsWebForGet(config),
 		bindings: Array.isArray(config.bindings) ? config.bindings : [],
 		mcp: buildSafeMcpConfigForWeb(config)

package/dist/src/gateway/hono/lib/safe-voice-config.d.ts

@@ -1,3 +1,17 @@
+export declare function maskRealtimeVoiceConfigForWeb(voice: {
+    realtime?: {
+        omni?: {
+            apiKey?: string;
+        };
+    };
+} | undefined): unknown;
+export declare function mergeRealtimeVoiceConfigPatch(previous: {
+    realtime?: {
+        omni?: {
+            apiKey?: string;
+        };
+    };
+} | undefined, incoming: unknown): unknown;
 /** Mask STT api keys for GET `/api/config`. */
 export declare function maskSttConfigForWeb(stt: unknown): unknown;
 /** Mask TTS api keys for GET `/api/config`. */

package/dist/src/gateway/hono/lib/safe-voice-config.js

@@ -1,6 +1,34 @@
 import { maskSecretLength } from "./mask-secret-length.js";
 import { isMaskedApiKey } from "../../config-tools-web.js";
 //#region src/gateway/hono/lib/safe-voice-config.ts
+function maskRealtimeVoiceConfigForWeb(voice) {
+	if (!voice?.realtime?.omni) return voice;
+	return {
+		...voice,
+		realtime: {
+			...voice.realtime,
+			omni: maskProviderSlice(voice.realtime.omni)
+		}
+	};
+}
+function mergeRealtimeVoiceConfigPatch(previous, incoming) {
+	if (!incoming || typeof incoming !== "object" || Array.isArray(incoming)) return incoming;
+	const patch = incoming;
+	const realtime = patch.realtime;
+	const omni = realtime?.omni;
+	if (!omni || typeof omni !== "object" || Array.isArray(omni)) return incoming;
+	const apiKey = omni.apiKey === "" ? void 0 : mergeApiKeyField(omni.apiKey, previous?.realtime?.omni?.apiKey);
+	return {
+		...patch,
+		realtime: {
+			...realtime,
+			omni: {
+				...omni,
+				...omni.apiKey !== void 0 ? { apiKey } : {}
+			}
+		}
+	};
+}
 function maskProviderSlice(slice) {
 	const next = { ...slice };
 	if (typeof next.apiKey === "string" && next.apiKey.trim()) next.apiKey = maskSecretLength(next.apiKey);
@@ -64,4 +92,4 @@ function mergeTtsConfigPatch(previous, incoming) {
 	return mergeVoiceConfigPatch(previous, incoming);
 }
 //#endregion
-export { maskSttConfigForWeb, maskTtsConfigForWeb, mergeSttConfigPatch, mergeTtsConfigPatch };
+export { maskRealtimeVoiceConfigForWeb, maskSttConfigForWeb, maskTtsConfigForWeb, mergeRealtimeVoiceConfigPatch, mergeSttConfigPatch, mergeTtsConfigPatch };

package/dist/src/gateway/hono/routes/config-patch/misc.js

@@ -9,7 +9,7 @@ import { isMaskedSecretPatchValue } from "../../lib/mask-secret-length.js";
 import { applyToolsWebPatch } from "../../../config-tools-web.js";
 import { mergeTunnelConfigPatch } from "../../../../tunnel/tunnel-config.js";
 import { assertGatewayRuntimeConfig } from "../../../runtime-config.js";
-import { mergeSttConfigPatch, mergeTtsConfigPatch } from "../../lib/safe-voice-config.js";
+import { mergeRealtimeVoiceConfigPatch, mergeSttConfigPatch, mergeTtsConfigPatch } from "../../lib/safe-voice-config.js";
 import { PATCH_OK, patchError } from "./result.js";
 //#region src/gateway/hono/routes/config-patch/misc.ts
 init_schema();
@@ -137,7 +137,7 @@ async function applyMiscPatch(config, body) {
 		config.messages.tts = mergeTtsConfigPatch(config.messages.tts, body.tts);
 	}
 	if (body.voice !== void 0) {
-		const parsed = VoiceConfigSchema.safeParse(body.voice);
+		const parsed = VoiceConfigSchema.safeParse(mergeRealtimeVoiceConfigPatch(config.voice, body.voice));
 		if (!parsed.success) return patchError(parsed.error.issues.map((i) => i.message).join("; "));
 		config.voice = parsed.data;
 	}

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

@@ -86,6 +86,13 @@ function registerFilesRoutes(authenticated, deps) {
 			return errorResponse(c, error);
 		}
 	});
+	authenticated.get("/api/files/spaces/:id/root", async (c) => {
+		try {
+			return c.json({ resource: await files.rootResource(c.req.param("id")) });
+		} catch (error) {
+			return errorResponse(c, error);
+		}
+	});
 	authenticated.get("/api/files/recent", async (c) => {
 		try {
 			const limit = Math.min(100, Math.max(1, Number(c.req.query("limit")) || 50));

package/dist/src/gateway/hono/routes/lazy-fallback.js

@@ -3,6 +3,16 @@ import { APP_LAZY_ROUTE_BUNDLES, AUTHENTICATED_LAZY_ROUTE_BUNDLES, findAuthentic
 import { Hono } from "hono";
 //#region src/gateway/hono/routes/lazy-fallback.ts
 const log = createGatewayRouteLogger("LazyRoutes");
+const FORWARDED_VARIABLES = Symbol("gateway.forwardedVariables");
+function createSubApp() {
+	const sub = new Hono();
+	sub.use(async (c, next) => {
+		const variables = c.env?.[FORWARDED_VARIABLES];
+		for (const [key, value] of Object.entries(variables ?? {})) c.set(key, value);
+		await next();
+	});
+	return sub;
+}
 const authenticatedSubApps = /* @__PURE__ */ new Map();
 const appSubApps = /* @__PURE__ */ new Map();
 const authenticatedLoadPromises = /* @__PURE__ */ new Map();
@@ -28,7 +38,7 @@ async function ensureAuthenticatedLazyBundle(bundleId, deps) {
 		if (!bundle) return null;
 		pending = (async () => {
 			const mod = await bundle.load();
-			const sub = new Hono();
+			const sub = createSubApp();
 			mod.register(sub, deps);
 			authenticatedSubApps.set(bundleId, sub);
 			authenticatedLoadPromises.delete(bundleId);
@@ -47,7 +57,7 @@ async function ensureAppLazyBundle(bundleId, params) {
 		if (!bundle) return null;
 		pending = (async () => {
 			const mod = await bundle.load();
-			const sub = new Hono();
+			const sub = createSubApp();
 			if (mod.registerOnApp) mod.registerOnApp(sub, params.service);
 			appSubApps.set(bundleId, sub);
 			appLoadPromises.delete(bundleId);
@@ -58,10 +68,14 @@ async function ensureAppLazyBundle(bundleId, params) {
 	return pending;
 }
 async function forwardToSubApp(c, sub) {
+	const env = {
+		...c.env,
+		[FORWARDED_VARIABLES]: { ...c.var }
+	};
 	try {
-		return await sub.fetch(c.req.raw, c.env, c.executionCtx);
+		return await sub.fetch(c.req.raw, env, c.executionCtx);
 	} catch (error) {
-		if (error instanceof Error && error.message.includes("ExecutionContext")) return sub.fetch(c.req.raw, c.env);
+		if (error instanceof Error && error.message.includes("ExecutionContext")) return sub.fetch(c.req.raw, env);
 		logRouteError(log, c, error, "gateway.route.lazy_forward");
 		throw error;
 	}

package/dist/src/gateway/hono/routes/runtime-tools.js

@@ -1,5 +1,6 @@
 import { init_paths_state, resolveStateDir } from "../../../config/paths-state.js";
 import { RuntimeToolsConfigSchema, init_schema } from "../../../config/schema.js";
+import { normalizeRuntimeVersionRequest } from "../../../runtime-tools/probe.js";
 import { ManagedRuntimeManager } from "../../../runtime-tools/manager.js";
 import { createGatewayRouteLogger } from "../lib/route-logger.js";
 import { pruneRuntimeTools } from "../../../runtime-tools/prune.js";
@@ -82,6 +83,11 @@ function registerRuntimeToolsRoutes(authenticated, deps) {
 			}, 400);
 			const body = await readJson(c);
 			const version = typeof body.version === "string" && body.version.trim() ? body.version.trim() : void 0;
+			if (version && !normalizeRuntimeVersionRequest(version)) return c.json({
+				ok: false,
+				error: "invalid-version",
+				message: "Invalid runtime version"
+			}, 400);
 			return streamSSE(c, async (stream) => {
 				const manager = runtimeManager(service.currentConfig);
 				let writes = Promise.resolve();
@@ -97,7 +103,7 @@ function registerRuntimeToolsRoutes(authenticated, deps) {
 				};
 				manager.on("progress", onProgress);
 				try {
-					const resolved = action === "repair" ? await manager.repair(runtime) : await manager.install(runtime, version);
+					const resolved = action === "repair" ? await manager.repair(runtime, version) : await manager.install(runtime, version);
 					await writes;
 					await stream.writeSSE({
 						event: "result",

package/dist/src/gateway/hono/routes/session-input-handler.js

@@ -73,6 +73,13 @@ async function submitSessionInput(c, deps, sessionKey) {
 		}
 	}, 400);
 	return withModelConfigLock(sessionKey, async () => {
+		if (deps.service.voiceRealtime?.hasConversation(sessionKey)) return c.json({
+			ok: false,
+			error: {
+				code: "SESSION_BUSY",
+				message: "End the voice call before sending text"
+			}
+		}, 409);
 		const selection = body.configVersion !== void 0 ? await deps.service.sessions.getAgentConfig(sessionKey) : void 0;
 		if (selection) {
 			if (!Number.isSafeInteger(body.configVersion) || selection.configVersion !== body.configVersion || !selection.fixedModel) return c.json({
@@ -181,6 +188,13 @@ async function replaceLatestSessionTurn(c, deps, sessionKey, targetTurnId) {
 		}
 	}, 401);
 	return withModelConfigLock(sessionKey, async () => {
+		if (deps.service.voiceRealtime?.hasConversation(sessionKey)) return c.json({
+			ok: false,
+			error: {
+				code: "SESSION_BUSY",
+				message: "End the voice call before editing a turn"
+			}
+		}, 409);
 		const selection = body.configVersion !== void 0 ? await deps.service.sessions.getAgentConfig(sessionKey) : void 0;
 		if (selection) {
 			if (!Number.isSafeInteger(body.configVersion) || selection.configVersion !== body.configVersion || !selection.fixedModel) return c.json({

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

@@ -3,13 +3,13 @@ import { resolveProjectAgentId } from "../../../projects/project-agent.js";
 import "../../../projects/index.js";
 import { evictEmbeddedSessionRunner } from "../../../agent/embedded/session-runner.js";
 import { messagesToClientHistory } from "../../../session/client-history.js";
-import { SessionEnvironmentService } from "../../../execution-environments/session-environment-service.js";
 import { deleteMediaUrisNoLongerReferenced } from "../../../media/session-references.js";
 import { getGatewayPrincipal } from "../../security/gateway-principal.js";
 import { createGatewayRouteLogger, logRouteError } from "../lib/route-logger.js";
 import { patchChatModelConfig } from "./chat-model-config.js";
 import { getSessionContextSummary } from "../../session-context-summary.js";
 import { respondStartupUnavailable } from "../lib/startup-unavailable.js";
+import { SessionEnvironmentService } from "../../../execution-environments/session-environment-service.js";
 import { randomUUID } from "node:crypto";
 //#region src/gateway/hono/routes/sessions.ts
 init_session_key();
@@ -785,6 +785,7 @@ function registerSessionsRoutes(authenticated, deps) {
 	});
 	authenticated.delete("/api/sessions/:key/messages", async (c) => {
 		const key = c.req.param("key");
+		if (service.voiceRealtime?.hasConversation(key)) return c.json({ error: "End the voice call before editing its history" }, 409);
 		const body = await c.req.json().catch(() => ({}));
 		const userRoundIndex = typeof body.userRoundIndex === "number" ? body.userRoundIndex : void 0;
 		if (userRoundIndex === void 0 || !Number.isInteger(userRoundIndex) || userRoundIndex < 0) return c.json({ error: "userRoundIndex must be a non-negative integer" }, 400);
@@ -805,6 +806,7 @@ function registerSessionsRoutes(authenticated, deps) {
 		const blocked = ensureGatewayReadyForSessions(c, service, "sessions.reset");
 		if (blocked) return blocked;
 		const key = c.req.param("key");
+		if (service.voiceRealtime?.hasConversation(key)) return c.json({ error: "End the voice call before resetting the session" }, 409);
 		const result = await service.sessions.reset(key);
 		if (result.ok === false) {
 			const status = result.error === "Session not found" ? 404 : 400;
@@ -824,6 +826,7 @@ function registerSessionsRoutes(authenticated, deps) {
 	});
 	authenticated.delete("/api/sessions/:key", async (c) => {
 		const key = c.req.param("key");
+		if (service.voiceRealtime?.hasConversation(key)) return c.json({ error: "End the voice call before deleting the session" }, 409);
 		if (environments.get(key) && service.getActiveWebchatRunId(key)) return c.json({
 			ok: false,
 			error: "Stop the active session run before deleting its environment"

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

@@ -2,7 +2,7 @@ import { getDefaultModelSync, init_providers, resolveModel } from "../../../prov
 import { completeWithResolvedCredentials } from "../../../providers/model-call.js";
 import { resolveTtsProviderConfigSlice } from "../../../voice/tts/config-slice.js";
 import { isTTSAvailable, resolveSpeechProvider } from "../../../voice/tts/factory.js";
-import { speak } from "../../../voice/tts/speak-core.js";
+import { speak, speakStream } from "../../../voice/tts/speak-core.js";
 import { mergeTtsConfigFromAppConfig } from "../../../voice/tts/merge-config.js";
 import { getLocalVoiceRuntimeClient } from "../../../voice/local/runtime-client.js";
 import { getAudioDecoderStatus } from "../../../voice/audio/normalize.js";
@@ -13,9 +13,14 @@ import { mergeSttConfigFromAppConfig } from "../../../voice/stt/config.js";
 import "../../../voice/stt/index.js";
 import { listSttProvidersForApi } from "../../../voice/stt/list-providers.js";
 import "../../../voice/tts/index.js";
+import { withModelConfigLock } from "../../../session/model-config-lock.js";
+import { createVoiceSessionRequestSchema } from "../../../../packages/realtime-protocol/src/voice.js";
+import { resolveOmniRoute } from "../../../voice/realtime/omniRoute.js";
+import { VoiceSessionCreationError, resolveStreamingStt, resolveStreamingTts } from "../../../voice/realtime/runtime.js";
+import { getGatewayPrincipal } from "../../security/gateway-principal.js";
 import { createGatewayRouteLogger } from "../lib/route-logger.js";
-import { listTtsProvidersForApi } from "../../../voice/tts/list-providers.js";
 import { listLocalVoiceModelStatuses, removeLocalVoiceModel, startLocalVoiceModelInstall } from "../../../voice/local/model-manager.js";
+import { listTtsProvidersForApi } from "../../../voice/tts/list-providers.js";
 //#region src/gateway/hono/routes/voice.ts
 init_providers();
 const log = createGatewayRouteLogger("Voice");
@@ -107,6 +112,134 @@ async function refineTranscript(raw, config, signal) {
 }
 function registerVoiceRoutes(authenticated, deps) {
 	const { service, strictRateLimitMiddleware } = deps;
+	authenticated.get("/api/voice/realtime/status", async (c) => {
+		const config = service.currentConfig;
+		const stt = resolveStreamingStt(config);
+		const tts = resolveStreamingTts(config);
+		const omni = await resolveOmniRoute(config).catch(() => null);
+		const enabled = config.voice?.realtime?.enabled === true;
+		const availability = (configured) => ({
+			available: enabled && configured,
+			...!enabled ? { reasonCode: "VOICE_DISABLED" } : !configured ? { reasonCode: "PROVIDER_UNAVAILABLE" } : {}
+		});
+		return c.json({
+			ok: true,
+			payload: {
+				enabled,
+				capabilities: {
+					dictation: availability(Boolean(stt)),
+					agent: availability(Boolean(stt && tts)),
+					omni: availability(Boolean(omni)),
+					languages: ["zh", "en"],
+					bargeIn: config.voice?.realtime?.bargeIn ?? true
+				},
+				defaultEngine: config.voice?.realtime?.defaultEngine ?? "agent",
+				omni: omni?.route ?? null,
+				stt: stt?.route ?? null,
+				tts: tts ? {
+					...tts.route,
+					voice: tts.config.providers?.[tts.route.provider]?.voice ?? tts.provider.providerConfig.voice
+				} : null
+			}
+		});
+	});
+	authenticated.post("/api/voice/realtime/preview", strictRateLimitMiddleware, async (c) => {
+		const config = service.currentConfig;
+		const route = resolveStreamingTts(config);
+		if (!route) return c.json({
+			ok: false,
+			error: { message: "Realtime voice output is not configured" }
+		}, 503);
+		const controller = new AbortController();
+		const signal = AbortSignal.any([
+			c.req.raw.signal,
+			controller.signal,
+			AbortSignal.timeout(2e4)
+		]);
+		let result;
+		let reader;
+		try {
+			result = await speakStream(config.voice?.language === "en" ? "Hello, I am your assistant. You can talk to me naturally." : "你好，我是你的助手。现在可以和我自然对话了。", route.config, {
+				signal,
+				appConfig: config,
+				allowFallback: false,
+				parseDirectives: false
+			});
+			if (result.outputFormat !== "pcm") throw new Error("Realtime preview requires PCM audio");
+			reader = result.audioStream.getReader();
+			const chunks = [];
+			let size = 0;
+			while (true) {
+				signal.throwIfAborted();
+				const { value, done } = await reader.read();
+				if (done) break;
+				size += value.byteLength;
+				if (size > 96e4) throw new Error("Realtime preview audio exceeds the limit");
+				chunks.push(value);
+			}
+			if (size === 0 || size % 2 !== 0) throw new Error("Realtime preview returned invalid audio");
+			return c.json({
+				ok: true,
+				payload: {
+					audio: Buffer.concat(chunks).toString("base64"),
+					sampleRate: 24e3
+				}
+			});
+		} catch (error) {
+			log.warn({
+				err: error,
+				provider: route.route.provider
+			}, "Realtime voice preview failed");
+			return c.json({
+				ok: false,
+				error: { message: "Voice preview failed. Check the provider credentials and try again." }
+			}, 502);
+		} finally {
+			controller.abort();
+			await reader?.cancel().catch(() => void 0);
+			reader?.releaseLock();
+			await result?.release();
+		}
+	});
+	for (const action of ["preflight", "sessions"]) authenticated.post(`/api/voice/realtime/${action}`, strictRateLimitMiddleware, async (c) => {
+		const parsed = createVoiceSessionRequestSchema.safeParse(await c.req.json().catch(() => null));
+		if (!parsed.success) return c.json({
+			ok: false,
+			error: {
+				code: "INVALID_REQUEST",
+				message: "Invalid realtime voice session request"
+			}
+		}, 400);
+		try {
+			if (action === "preflight") {
+				await service.voiceRealtime.preflight(parsed.data);
+				return c.json({ ok: true });
+			}
+			const principal = getGatewayPrincipal(c);
+			const create = () => service.voiceRealtime.createSession(parsed.data, principal.principalId);
+			const payload = parsed.data.sessionKey ? await withModelConfigLock(parsed.data.sessionKey, create) : await create();
+			return c.json({
+				ok: true,
+				payload
+			});
+		} catch (error) {
+			const message = error instanceof Error ? error.message : "Realtime voice is unavailable";
+			if (error instanceof VoiceSessionCreationError) return c.json({
+				ok: false,
+				error: {
+					code: error.code,
+					message
+				}
+			}, error.status);
+			return c.json({
+				ok: false,
+				error: {
+					code: "PROVIDER_UNAVAILABLE",
+					message
+				}
+			}, 503);
+		}
+	});
 	authenticated.post("/api/voice/language", strictRateLimitMiddleware, async (c) => {
 		let body = {};
 		try {
@@ -153,7 +286,11 @@ function registerVoiceRoutes(authenticated, deps) {
 			error: { message: "provider and model are required" }
 		}, 400);
 		const config = service.currentConfig;
-		const baseConfig = mergeTtsConfigFromAppConfig(config.messages?.tts);
+		const baseConfig = c.req.query("purpose") === "realtime" ? resolveStreamingTts(config)?.config : mergeTtsConfigFromAppConfig(config.messages?.tts);
+		if (!baseConfig) return c.json({
+			ok: true,
+			payload: { voices: [] }
+		});
 		const resolved = resolveSpeechProvider(provider, {
 			...baseConfig,
 			provider,

package/dist/src/gateway/server.js

@@ -68,6 +68,7 @@ var GatewayServer = class {
 			});
 			inner.on("upgrade", (req, socket, head) => {
 				try {
+					if (this.service.voiceRealtime.handleUpgrade(req, socket, head)) return;
 					if (this.service.realtime.handleUpgrade(req, socket, head)) return;
 					if (handleSiteShareUpgrade(this.service, req, socket, head)) return;
 					socket.destroy();

package/dist/src/gateway/service/agent-runner.d.ts

@@ -71,6 +71,7 @@ export declare class GatewayAgentRunner {
         signal?: AbortSignal;
         runId?: string;
         sourceContexts?: AgentSourceContext[];
+        presentation?: 'voice';
     }): AsyncGenerator<{
         type: string;
         [key: string]: unknown;

package/dist/src/gateway/service/agent-runner.js

@@ -128,12 +128,7 @@ var GatewayAgentRunner = class {
 			completeRealtimeTopic: this.opts.completeRealtimeTopic
 		}, message, channel, chatId, origin, attachments, thinking, runOptions);
 		try {
-			let step = await iter.next();
-			while (!step.done) {
-				yield step.value;
-				step = await iter.next();
-			}
-			return step.value;
+			return yield* iter;
 		} finally {
 			if (trackedRunId) {
 				this.resolveRunCompletions.get(trackedRunId)?.();

package/dist/src/gateway/service/run-gateway-agent.d.ts

@@ -25,6 +25,7 @@ export declare function runGatewayAgent(deps: RunGatewayAgentDeps, message: stri
     signal?: AbortSignal;
     runId?: string;
     sourceContexts?: AgentSourceContext[];
+    presentation?: 'voice';
 }): AsyncGenerator<RunGatewayAgentYield, {
     status: string;
     summary: string;