Vendor OpenClaw source as Adolf fork baseline

Adolf is a fork/vendored clone of github.com/openclaw/openclaw (v2026.6.11),
free to diverge. Tree copied sans upstream .git; upstream remote added for
future syncs. Node pinned to 24 (.nvmrc); engines already require >=22.19.
Preserves docs/ARCHITECTURE.md.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LeqyaxJF2nbRXJtae2kNB2

extensions/microsoft/index.ts

@@ -0,0 +1,12 @@
+// Microsoft plugin entrypoint registers its OpenClaw integration.
+import { definePluginEntry } from "openclaw/plugin-sdk/plugin-entry";
+import { buildMicrosoftSpeechProvider } from "./speech-provider.js";
+
+export default definePluginEntry({
+  id: "microsoft",
+  name: "Microsoft Speech",
+  description: "Bundled Microsoft speech provider",
+  register(api) {
+    api.registerSpeechProvider(buildMicrosoftSpeechProvider());
+  },
+});

extensions/microsoft/microsoft.live.test.ts

@@ -0,0 +1,15 @@
+// Microsoft tests cover microsoft plugin behavior.
+import { isLiveTestEnabled } from "openclaw/plugin-sdk/test-env";
+import { describe, expect, it } from "vitest";
+import { listMicrosoftVoices } from "./speech-provider.js";
+
+const describeLive = isLiveTestEnabled() ? describe : describe.skip;
+
+describeLive("microsoft plugin live", () => {
+  it("lists Edge speech voices", async () => {
+    const voices = await listMicrosoftVoices();
+
+    expect(voices.length).toBeGreaterThan(100);
+    expect(voices.map((voice) => voice.id)).toContain("en-US-MichelleNeural");
+  }, 60_000);
+});

extensions/microsoft/openclaw.plugin.json

@@ -0,0 +1,15 @@
+{
+  "id": "microsoft",
+  "activation": {
+    "onStartup": false
+  },
+  "enabledByDefault": true,
+  "contracts": {
+    "speechProviders": ["microsoft"]
+  },
+  "configSchema": {
+    "type": "object",
+    "additionalProperties": false,
+    "properties": {}
+  }
+}

extensions/microsoft/package.json

@@ -0,0 +1,18 @@
+{
+  "name": "@openclaw/microsoft-speech",
+  "version": "2026.6.11",
+  "private": true,
+  "description": "OpenClaw Microsoft speech plugin",
+  "type": "module",
+  "dependencies": {
+    "node-edge-tts": "1.2.10"
+  },
+  "devDependencies": {
+    "@openclaw/plugin-sdk": "workspace:*"
+  },
+  "openclaw": {
+    "extensions": [
+      "./index.ts"
+    ]
+  }
+}

extensions/microsoft/speech-provider.test.ts

@@ -0,0 +1,287 @@
+// Microsoft tests cover speech provider plugin behavior.
+import { mkdtempSync, writeFileSync } from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import type { OpenClawConfig } from "openclaw/plugin-sdk/config-contracts";
+import {
+  finalizeDebugProxyCapture,
+  getDebugProxyCaptureStore,
+  initializeDebugProxyCapture,
+} from "openclaw/plugin-sdk/proxy-capture";
+import { afterEach, describe, expect, it, vi } from "vitest";
+import { installDebugProxyTestResetHooks } from "../test-support/debug-proxy-env-test-helpers.js";
+
+vi.mock("node-edge-tts", () => ({
+  EdgeTTS: class {
+    async ttsPromise(): Promise<void> {}
+  },
+}));
+
+import {
+  buildMicrosoftSpeechProvider,
+  isCjkDominant,
+  listMicrosoftVoices,
+} from "./speech-provider.js";
+import * as ttsModule from "./tts.js";
+
+const TEST_CFG = {} as OpenClawConfig;
+
+function requireFirstEdgeTtsCall(edgeSpy: ReturnType<typeof vi.spyOn>): {
+  config?: unknown;
+  outputPath: string;
+  text?: string;
+  timeoutMs?: number;
+} {
+  const [call] = edgeSpy.mock.calls;
+  if (!call) {
+    throw new Error("expected Microsoft Edge TTS call");
+  }
+  const [edgeCall] = call;
+  if (!edgeCall || typeof edgeCall !== "object" || Array.isArray(edgeCall)) {
+    throw new Error("expected Microsoft Edge TTS call");
+  }
+  return edgeCall as {
+    config?: unknown;
+    outputPath: string;
+    text?: string;
+    timeoutMs?: number;
+  };
+}
+
+describe("listMicrosoftVoices", () => {
+  const proxyReset = installDebugProxyTestResetHooks();
+
+  it("maps Microsoft voice metadata into speech voice options", async () => {
+    globalThis.fetch = vi.fn().mockResolvedValue(
+      new Response(
+        JSON.stringify([
+          {
+            ShortName: "en-US-AvaNeural",
+            FriendlyName: "Microsoft Ava Online (Natural) - English (United States)",
+            Locale: "en-US",
+            Gender: "Female",
+            VoiceTag: {
+              ContentCategories: ["General"],
+              VoicePersonalities: ["Friendly", "Positive"],
+            },
+          },
+        ]),
+        { status: 200 },
+      ),
+    ) as unknown as typeof globalThis.fetch;
+
+    const voices = await listMicrosoftVoices();
+
+    expect(voices).toEqual([
+      {
+        id: "en-US-AvaNeural",
+        name: "Microsoft Ava Online (Natural) - English (United States)",
+        category: "General",
+        description: "Friendly, Positive",
+        locale: "en-US",
+        gender: "Female",
+        personalities: ["Friendly", "Positive"],
+      },
+    ]);
+  });
+
+  it("throws on Microsoft voice list failures", async () => {
+    globalThis.fetch = vi
+      .fn()
+      .mockResolvedValue(
+        new Response("nope", { status: 503 }),
+      ) as unknown as typeof globalThis.fetch;
+
+    await expect(listMicrosoftVoices()).rejects.toThrow("Microsoft voices API error (503)");
+  });
+
+  it("records voice discovery exchanges in debug proxy capture mode", async () => {
+    const tempDir = mkdtempSync(path.join(os.tmpdir(), "microsoft-voices-capture-"));
+    proxyReset.captureProxyEnv();
+    process.env.OPENCLAW_DEBUG_PROXY_ENABLED = "1";
+    process.env.OPENCLAW_STATE_DIR = tempDir;
+    process.env.OPENCLAW_DEBUG_PROXY_SESSION_ID = "ms-voices-session";
+
+    globalThis.fetch = vi
+      .fn()
+      .mockResolvedValue(
+        new Response(JSON.stringify([{ ShortName: "en-US-AvaNeural" }]), { status: 200 }),
+      ) as unknown as typeof globalThis.fetch;
+
+    const store = getDebugProxyCaptureStore();
+    store.upsertSession({
+      id: "ms-voices-session",
+      startedAt: Date.now(),
+      mode: "test",
+      sourceScope: "openclaw",
+      sourceProcess: "openclaw",
+    });
+
+    await listMicrosoftVoices();
+
+    await vi.waitFor(() => {
+      const events = store.getSessionEvents("ms-voices-session", 10);
+      expect(
+        events.some(
+          (event) => event.kind === "request" && event.host === "speech.platform.bing.com",
+        ),
+      ).toBe(true);
+      expect(
+        events.some(
+          (event) => event.kind === "response" && event.host === "speech.platform.bing.com",
+        ),
+      ).toBe(true);
+    });
+  });
+
+  it("does not double-capture voice discovery when the global fetch patch is installed", async () => {
+    const tempDir = mkdtempSync(path.join(os.tmpdir(), "microsoft-voices-global-"));
+    proxyReset.captureProxyEnv();
+    process.env.OPENCLAW_DEBUG_PROXY_ENABLED = "1";
+    process.env.OPENCLAW_STATE_DIR = tempDir;
+    process.env.OPENCLAW_DEBUG_PROXY_SESSION_ID = "ms-voices-global-session";
+
+    globalThis.fetch = vi.fn(
+      async () => new Response(JSON.stringify([{ ShortName: "en-US-AvaNeural" }]), { status: 200 }),
+    ) as unknown as typeof globalThis.fetch;
+
+    const store = getDebugProxyCaptureStore();
+    store.upsertSession({
+      id: "ms-voices-global-session",
+      startedAt: Date.now(),
+      mode: "test",
+      sourceScope: "openclaw",
+      sourceProcess: "openclaw",
+    });
+    initializeDebugProxyCapture("test");
+
+    try {
+      await listMicrosoftVoices();
+
+      let events: Array<Record<string, unknown>> = [];
+      await vi.waitFor(() => {
+        events = store
+          .getSessionEvents("ms-voices-global-session", 10)
+          .filter((event) => event.host === "speech.platform.bing.com");
+        expect(events).toHaveLength(2);
+      });
+      const kinds = events.map((event) => String(event.kind)).toSorted();
+      expect(kinds).toEqual(["request", "response"]);
+    } finally {
+      globalThis.fetch = proxyReset.originalFetch;
+      finalizeDebugProxyCapture();
+    }
+  });
+});
+
+describe("isCjkDominant", () => {
+  it("returns true for Chinese text", () => {
+    expect(isCjkDominant("你好世界")).toBe(true);
+  });
+
+  it("returns true for mixed text with majority CJK", () => {
+    expect(isCjkDominant("你好，这是一个测试 hello")).toBe(true);
+  });
+
+  it("returns false for English text", () => {
+    expect(isCjkDominant("Hello, this is a test")).toBe(false);
+  });
+
+  it("returns false for empty string", () => {
+    expect(isCjkDominant("")).toBe(false);
+  });
+
+  it("returns false for mostly English with a few CJK chars", () => {
+    expect(isCjkDominant("This is a long English sentence with one 字")).toBe(false);
+  });
+});
+
+describe("buildMicrosoftSpeechProvider", () => {
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  it("switches to a Chinese voice for CJK text when no explicit voice override is set", async () => {
+    const provider = buildMicrosoftSpeechProvider();
+    const edgeSpy = vi.spyOn(ttsModule, "edgeTTS").mockImplementation(async ({ outputPath }) => {
+      writeFileSync(outputPath, Buffer.from([0xff, 0xfb, 0x90, 0x00]));
+    });
+
+    await provider.synthesize({
+      text: "你好，这是一个测试 hello",
+      cfg: TEST_CFG,
+      providerConfig: {
+        enabled: true,
+        voice: "en-US-MichelleNeural",
+        lang: "en-US",
+        outputFormat: "audio-24khz-48kbitrate-mono-mp3",
+        outputFormatConfigured: true,
+        saveSubtitles: false,
+      },
+      providerOverrides: {},
+      timeoutMs: 1000,
+      target: "audio-file",
+    });
+
+    expect(edgeSpy).toHaveBeenCalledOnce();
+    const edgeCall = requireFirstEdgeTtsCall(edgeSpy);
+    expect(edgeCall.text).toBe("你好，这是一个测试 hello");
+    expect(path.basename(edgeCall.outputPath)).toBe("speech.mp3");
+    expect(edgeCall.timeoutMs).toBe(1000);
+    expect(edgeCall.config).toEqual({
+      enabled: true,
+      voice: "zh-CN-XiaoxiaoNeural",
+      lang: "zh-CN",
+      outputFormat: "audio-24khz-48kbitrate-mono-mp3",
+      outputFormatConfigured: true,
+      pitch: undefined,
+      rate: undefined,
+      volume: undefined,
+      saveSubtitles: false,
+      proxy: undefined,
+      timeoutMs: undefined,
+    });
+  });
+
+  it("preserves an explicitly configured English voice for CJK text", async () => {
+    const provider = buildMicrosoftSpeechProvider();
+    const edgeSpy = vi.spyOn(ttsModule, "edgeTTS").mockImplementation(async ({ outputPath }) => {
+      writeFileSync(outputPath, Buffer.from([0xff, 0xfb, 0x90, 0x00]));
+    });
+
+    await provider.synthesize({
+      text: "你好，这是一个测试 hello",
+      cfg: TEST_CFG,
+      providerConfig: {
+        enabled: true,
+        voice: "en-US-AvaNeural",
+        lang: "en-US",
+        outputFormat: "audio-24khz-48kbitrate-mono-mp3",
+        outputFormatConfigured: true,
+        saveSubtitles: false,
+      },
+      providerOverrides: {},
+      timeoutMs: 1000,
+      target: "audio-file",
+    });
+
+    expect(edgeSpy).toHaveBeenCalledOnce();
+    const edgeCall = requireFirstEdgeTtsCall(edgeSpy);
+    expect(edgeCall.text).toBe("你好，这是一个测试 hello");
+    expect(path.basename(edgeCall.outputPath)).toBe("speech.mp3");
+    expect(edgeCall.timeoutMs).toBe(1000);
+    expect(edgeCall.config).toEqual({
+      enabled: true,
+      voice: "en-US-AvaNeural",
+      lang: "en-US",
+      outputFormat: "audio-24khz-48kbitrate-mono-mp3",
+      outputFormatConfigured: true,
+      pitch: undefined,
+      rate: undefined,
+      volume: undefined,
+      saveSubtitles: false,
+      proxy: undefined,
+      timeoutMs: undefined,
+    });
+  });
+});

extensions/microsoft/speech-provider.ts

@@ -0,0 +1,302 @@
+// Microsoft provider module implements model/runtime integration.
+import { readFileSync } from "node:fs";
+import path from "node:path";
+import {
+  CHROMIUM_FULL_VERSION,
+  TRUSTED_CLIENT_TOKEN,
+  generateSecMsGecToken,
+} from "node-edge-tts/dist/drm.js";
+import { isVoiceCompatibleAudio } from "openclaw/plugin-sdk/media-runtime";
+import {
+  assertOkOrThrowProviderError,
+  readProviderJsonResponse,
+} from "openclaw/plugin-sdk/provider-http";
+import {
+  captureHttpExchange,
+  isDebugProxyGlobalFetchPatchInstalled,
+} from "openclaw/plugin-sdk/proxy-capture";
+import type {
+  SpeechProviderConfig,
+  SpeechProviderPlugin,
+  SpeechVoiceOption,
+} from "openclaw/plugin-sdk/speech";
+import { asBoolean, asFiniteNumber, asObject, trimToUndefined } from "openclaw/plugin-sdk/speech";
+import {
+  fetchWithSsrFGuard,
+  ssrfPolicyFromHttpBaseUrlAllowedHostname,
+} from "openclaw/plugin-sdk/ssrf-runtime";
+import { tempWorkspace, resolvePreferredOpenClawTmpDir } from "openclaw/plugin-sdk/temp-path";
+import { edgeTTS, inferEdgeExtension } from "./tts.js";
+
+const DEFAULT_EDGE_VOICE = "en-US-MichelleNeural";
+const DEFAULT_EDGE_LANG = "en-US";
+const DEFAULT_EDGE_OUTPUT_FORMAT = "audio-24khz-48kbitrate-mono-mp3";
+
+type MicrosoftProviderConfig = {
+  enabled: boolean;
+  voice: string;
+  lang: string;
+  outputFormat: string;
+  outputFormatConfigured: boolean;
+  pitch?: string;
+  rate?: string;
+  volume?: string;
+  saveSubtitles: boolean;
+  proxy?: string;
+  timeoutMs?: number;
+};
+
+type MicrosoftVoiceListEntry = {
+  ShortName?: string;
+  FriendlyName?: string;
+  Locale?: string;
+  Gender?: string;
+  VoiceTag?: {
+    ContentCategories?: string[];
+    VoicePersonalities?: string[];
+  };
+};
+
+function normalizeMicrosoftProviderConfig(
+  rawConfig: Record<string, unknown>,
+): MicrosoftProviderConfig {
+  const providers = asObject(rawConfig.providers);
+  const rawEdge = asObject(rawConfig.edge);
+  const rawMicrosoft = asObject(rawConfig.microsoft);
+  const rawProviderMicrosoft = asObject(providers?.microsoft);
+  const raw = { ...rawEdge, ...rawMicrosoft, ...rawProviderMicrosoft };
+  const outputFormat = trimToUndefined(raw.outputFormat);
+  return {
+    enabled: asBoolean(raw.enabled) ?? true,
+    voice: trimToUndefined(raw.voice) ?? DEFAULT_EDGE_VOICE,
+    lang: trimToUndefined(raw.lang) ?? DEFAULT_EDGE_LANG,
+    outputFormat: outputFormat ?? DEFAULT_EDGE_OUTPUT_FORMAT,
+    outputFormatConfigured: Boolean(outputFormat),
+    pitch: trimToUndefined(raw.pitch),
+    rate: trimToUndefined(raw.rate),
+    volume: trimToUndefined(raw.volume),
+    saveSubtitles: asBoolean(raw.saveSubtitles) ?? false,
+    proxy: trimToUndefined(raw.proxy),
+    timeoutMs: asFiniteNumber(raw.timeoutMs),
+  };
+}
+
+function readMicrosoftProviderConfig(config: SpeechProviderConfig): MicrosoftProviderConfig {
+  const defaults = normalizeMicrosoftProviderConfig({});
+  return {
+    enabled: asBoolean(config.enabled) ?? defaults.enabled,
+    voice: trimToUndefined(config.voice) ?? defaults.voice,
+    lang: trimToUndefined(config.lang) ?? defaults.lang,
+    outputFormat: trimToUndefined(config.outputFormat) ?? defaults.outputFormat,
+    outputFormatConfigured:
+      asBoolean(config.outputFormatConfigured) ?? defaults.outputFormatConfigured,
+    pitch: trimToUndefined(config.pitch) ?? defaults.pitch,
+    rate: trimToUndefined(config.rate) ?? defaults.rate,
+    volume: trimToUndefined(config.volume) ?? defaults.volume,
+    saveSubtitles: asBoolean(config.saveSubtitles) ?? defaults.saveSubtitles,
+    proxy: trimToUndefined(config.proxy) ?? defaults.proxy,
+    timeoutMs: asFiniteNumber(config.timeoutMs) ?? defaults.timeoutMs,
+  };
+}
+
+function buildMicrosoftVoiceHeaders(): Record<string, string> {
+  const major = CHROMIUM_FULL_VERSION.split(".")[0] || "0";
+  return {
+    Authority: "speech.platform.bing.com",
+    Origin: "chrome-extension://jdiccldimpdaibmpdkjnbmckianbfold",
+    Accept: "*/*",
+    "User-Agent":
+      `Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 ` +
+      `(KHTML, like Gecko) Chrome/${major}.0.0.0 Safari/537.36 Edg/${major}.0.0.0`,
+    "Sec-MS-GEC": generateSecMsGecToken(),
+    "Sec-MS-GEC-Version": `1-${CHROMIUM_FULL_VERSION}`,
+  };
+}
+
+function formatMicrosoftVoiceDescription(entry: MicrosoftVoiceListEntry): string | undefined {
+  const personalities = entry.VoiceTag?.VoicePersonalities?.filter(Boolean) ?? [];
+  return personalities.length > 0 ? personalities.join(", ") : undefined;
+}
+
+export function isCjkDominant(text: string): boolean {
+  const stripped = text.replace(/\s+/g, "");
+  if (stripped.length === 0) {
+    return false;
+  }
+  let cjkCount = 0;
+  for (const ch of stripped) {
+    const code = ch.codePointAt(0) ?? 0;
+    if (
+      (code >= 0x4e00 && code <= 0x9fff) ||
+      (code >= 0x3400 && code <= 0x4dbf) ||
+      (code >= 0x3000 && code <= 0x303f) ||
+      (code >= 0xff00 && code <= 0xffef)
+    ) {
+      cjkCount += 1;
+    }
+  }
+  return cjkCount / stripped.length > 0.3;
+}
+
+const DEFAULT_CHINESE_EDGE_VOICE = "zh-CN-XiaoxiaoNeural";
+const DEFAULT_CHINESE_EDGE_LANG = "zh-CN";
+
+export async function listMicrosoftVoices(): Promise<SpeechVoiceOption[]> {
+  const url =
+    "https://speech.platform.bing.com/consumer/speech/synthesize/readaloud/voices/list" +
+    `?trustedclienttoken=${TRUSTED_CLIENT_TOKEN}`;
+  const headers = buildMicrosoftVoiceHeaders();
+  const { response, release } = await fetchWithSsrFGuard({
+    url,
+    init: {
+      headers,
+    },
+    policy: ssrfPolicyFromHttpBaseUrlAllowedHostname("https://speech.platform.bing.com"),
+    auditContext: "microsoft.speech.voices",
+  });
+  try {
+    if (!isDebugProxyGlobalFetchPatchInstalled()) {
+      captureHttpExchange({
+        url,
+        method: "GET",
+        requestHeaders: headers,
+        response,
+        transport: "http",
+        meta: {
+          provider: "microsoft",
+          capability: "speech-voices",
+        },
+      });
+    }
+    await assertOkOrThrowProviderError(response, "Microsoft voices API error");
+    const voices = await readProviderJsonResponse<MicrosoftVoiceListEntry[]>(
+      response,
+      "microsoft.speech-voices",
+    );
+    return Array.isArray(voices)
+      ? voices
+          .map((voice) => ({
+            id: voice.ShortName?.trim() ?? "",
+            name: trimToUndefined(voice.FriendlyName) ?? trimToUndefined(voice.ShortName),
+            category: voice.VoiceTag?.ContentCategories?.find((value) => value.trim().length > 0),
+            description: formatMicrosoftVoiceDescription(voice),
+            locale: trimToUndefined(voice.Locale),
+            gender: trimToUndefined(voice.Gender),
+            personalities: voice.VoiceTag?.VoicePersonalities?.filter(
+              (value): value is string => value.trim().length > 0,
+            ),
+          }))
+          .filter((voice) => voice.id.length > 0)
+      : [];
+  } finally {
+    await release();
+  }
+}
+
+export function buildMicrosoftSpeechProvider(): SpeechProviderPlugin {
+  return {
+    id: "microsoft",
+    label: "Microsoft",
+    aliases: ["edge"],
+    autoSelectOrder: 30,
+    resolveConfig: ({ rawConfig }) => normalizeMicrosoftProviderConfig(rawConfig),
+    resolveTalkConfig: ({ baseTtsConfig, talkProviderConfig }) => {
+      const base = normalizeMicrosoftProviderConfig(baseTtsConfig);
+      return {
+        ...base,
+        enabled: true,
+        ...(trimToUndefined(talkProviderConfig.voiceId) == null
+          ? {}
+          : { voice: trimToUndefined(talkProviderConfig.voiceId) }),
+        ...(trimToUndefined(talkProviderConfig.languageCode) == null
+          ? {}
+          : { lang: trimToUndefined(talkProviderConfig.languageCode) }),
+        ...(trimToUndefined(talkProviderConfig.outputFormat) == null
+          ? {}
+          : { outputFormat: trimToUndefined(talkProviderConfig.outputFormat) }),
+        ...(trimToUndefined(talkProviderConfig.pitch) == null
+          ? {}
+          : { pitch: trimToUndefined(talkProviderConfig.pitch) }),
+        ...(trimToUndefined(talkProviderConfig.rate) == null
+          ? {}
+          : { rate: trimToUndefined(talkProviderConfig.rate) }),
+        ...(trimToUndefined(talkProviderConfig.volume) == null
+          ? {}
+          : { volume: trimToUndefined(talkProviderConfig.volume) }),
+        ...(trimToUndefined(talkProviderConfig.proxy) == null
+          ? {}
+          : { proxy: trimToUndefined(talkProviderConfig.proxy) }),
+        ...(asFiniteNumber(talkProviderConfig.timeoutMs) == null
+          ? {}
+          : { timeoutMs: asFiniteNumber(talkProviderConfig.timeoutMs) }),
+      };
+    },
+    resolveTalkOverrides: ({ params }) => ({
+      ...(trimToUndefined(params.voiceId) == null
+        ? {}
+        : { voice: trimToUndefined(params.voiceId) }),
+      ...(trimToUndefined(params.outputFormat) == null
+        ? {}
+        : { outputFormat: trimToUndefined(params.outputFormat) }),
+    }),
+    listVoices: async () => await listMicrosoftVoices(),
+    isConfigured: ({ providerConfig }) => readMicrosoftProviderConfig(providerConfig).enabled,
+    synthesize: async (req) => {
+      const config = readMicrosoftProviderConfig(req.providerConfig);
+      const temp = await tempWorkspace({
+        rootDir: resolvePreferredOpenClawTmpDir(),
+        prefix: "tts-microsoft-",
+      });
+      const tempDir = temp.dir;
+      const overrideVoice = trimToUndefined(req.providerOverrides?.voice);
+      let voice = overrideVoice ?? config.voice;
+      let lang = config.lang;
+      let outputFormat =
+        trimToUndefined(req.providerOverrides?.outputFormat) ?? config.outputFormat;
+      const fallbackOutputFormat =
+        outputFormat !== DEFAULT_EDGE_OUTPUT_FORMAT ? DEFAULT_EDGE_OUTPUT_FORMAT : undefined;
+
+      if (!overrideVoice && voice === DEFAULT_EDGE_VOICE && isCjkDominant(req.text)) {
+        voice = DEFAULT_CHINESE_EDGE_VOICE;
+        lang = DEFAULT_CHINESE_EDGE_LANG;
+      }
+
+      try {
+        const runEdge = async (format: string) => {
+          const fileExtension = inferEdgeExtension(format);
+          const outputPath = path.join(tempDir, `speech${fileExtension}`);
+          await edgeTTS({
+            text: req.text,
+            outputPath,
+            config: {
+              ...config,
+              voice,
+              lang,
+              outputFormat: format,
+            },
+            timeoutMs: req.timeoutMs,
+          });
+          const audioBuffer = readFileSync(outputPath);
+          return {
+            audioBuffer,
+            outputFormat: format,
+            fileExtension,
+            voiceCompatible: isVoiceCompatibleAudio({ fileName: outputPath }),
+          };
+        };
+
+        try {
+          return await runEdge(outputFormat);
+        } catch (error) {
+          if (!fallbackOutputFormat || fallbackOutputFormat === outputFormat) {
+            throw error;
+          }
+          outputFormat = fallbackOutputFormat;
+          return await runEdge(outputFormat);
+        }
+      } finally {
+        await temp.cleanup();
+      }
+    },
+  };
+}

extensions/microsoft/test-api.ts

@@ -0,0 +1,2 @@
+// Microsoft API module exposes the plugin public contract.
+export { buildMicrosoftSpeechProvider } from "./speech-provider.js";

extensions/microsoft/tsconfig.json

@@ -0,0 +1,16 @@
+{
+  "extends": "../tsconfig.package-boundary.base.json",
+  "compilerOptions": {
+    "rootDir": "."
+  },
+  "include": ["./*.ts", "./src/**/*.ts"],
+  "exclude": [
+    "./**/*.test.ts",
+    "./dist/**",
+    "./node_modules/**",
+    "./src/test-support/**",
+    "./src/**/*test-helpers.ts",
+    "./src/**/*test-harness.ts",
+    "./src/**/*test-support.ts"
+  ]
+}

extensions/microsoft/tts.test.ts

@@ -0,0 +1,194 @@
+// Microsoft tests cover tts plugin behavior.
+import { existsSync, mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import path from "node:path";
+import { afterEach, beforeAll, describe, expect, it, vi } from "vitest";
+
+let edgeTTS: typeof import("./tts.js").edgeTTS;
+
+function createEdgeTTSDeps(
+  ttsPromise: (text: string, filePath: string) => Promise<void>,
+  onConstruct?: () => void,
+) {
+  return {
+    EdgeTTS: class {
+      constructor() {
+        onConstruct?.();
+      }
+
+      ttsPromise(text: string, filePath: string) {
+        return ttsPromise(text, filePath);
+      }
+    },
+  };
+}
+
+const baseEdgeConfig = {
+  voice: "en-US-MichelleNeural",
+  lang: "en-US",
+  outputFormat: "audio-24khz-48kbitrate-mono-mp3",
+  saveSubtitles: false,
+};
+
+describe("edgeTTS empty audio validation", () => {
+  let tempDir: string | undefined;
+
+  beforeAll(async () => {
+    ({ edgeTTS } = await import("./tts.js"));
+  });
+
+  afterEach(() => {
+    if (tempDir) {
+      rmSync(tempDir, { recursive: true, force: true });
+      tempDir = undefined;
+    }
+  });
+
+  it("rejects blank text before constructing Edge TTS", async () => {
+    tempDir = mkdtempSync(path.join(tmpdir(), "tts-test-"));
+    const outputPath = path.join(tempDir, "voice.mp3");
+    const onConstruct = vi.fn();
+    const deps = createEdgeTTSDeps(async (_text: string, filePath: string) => {
+      writeFileSync(filePath, Buffer.from([0xff]));
+    }, onConstruct);
+
+    await expect(
+      edgeTTS(
+        {
+          text: " \n\t ",
+          outputPath,
+          config: baseEdgeConfig,
+          timeoutMs: 10000,
+        },
+        deps,
+      ),
+    ).rejects.toThrow("Microsoft TTS text cannot be empty");
+    expect(onConstruct).not.toHaveBeenCalled();
+  });
+
+  it("throws after one retry when the output file stays empty", async () => {
+    tempDir = mkdtempSync(path.join(tmpdir(), "tts-test-"));
+    const outputPath = path.join(tempDir, "voice.mp3");
+    const calls: string[] = [];
+
+    const deps = createEdgeTTSDeps(async (text: string, filePath: string) => {
+      calls.push(text);
+      writeFileSync(filePath, "");
+    });
+
+    await expect(
+      edgeTTS(
+        {
+          text: "Hello",
+          outputPath,
+          config: baseEdgeConfig,
+          timeoutMs: 10000,
+        },
+        deps,
+      ),
+    ).rejects.toThrow("Edge TTS produced empty audio file after retry");
+    expect(calls).toEqual(["Hello", "Hello"]);
+  });
+
+  it("succeeds when the output file has content", async () => {
+    tempDir = mkdtempSync(path.join(tmpdir(), "tts-test-"));
+    const outputPath = path.join(tempDir, "voice.mp3");
+    let stagedPath = "";
+
+    const deps = createEdgeTTSDeps(async (_text: string, filePath: string) => {
+      stagedPath = filePath;
+      writeFileSync(filePath, Buffer.from([0xff, 0xfb, 0x90, 0x00]));
+    });
+
+    await expect(
+      edgeTTS(
+        {
+          text: "Hello",
+          outputPath,
+          config: baseEdgeConfig,
+          timeoutMs: 10000,
+        },
+        deps,
+      ),
+    ).resolves.toBeUndefined();
+    expect(stagedPath).not.toBe(outputPath);
+    expect(path.basename(stagedPath)).toContain(path.basename(outputPath));
+    expect(path.basename(stagedPath)).toMatch(/\.part$/);
+    expect(readFileSync(outputPath)).toEqual(Buffer.from([0xff, 0xfb, 0x90, 0x00]));
+    expect(existsSync(stagedPath)).toBe(false);
+  });
+
+  it("retries once when the first output file is empty", async () => {
+    tempDir = mkdtempSync(path.join(tmpdir(), "tts-test-"));
+    const outputPath = path.join(tempDir, "voice.mp3");
+    const calls: string[] = [];
+
+    const deps = createEdgeTTSDeps(async (text: string, filePath: string) => {
+      calls.push(text);
+      writeFileSync(filePath, calls.length === 1 ? "" : Buffer.from([0xff, 0xfb, 0x90, 0x00]));
+    });
+
+    await expect(
+      edgeTTS(
+        {
+          text: "Hello",
+          outputPath,
+          config: baseEdgeConfig,
+          timeoutMs: 10000,
+        },
+        deps,
+      ),
+    ).resolves.toBeUndefined();
+    expect(calls).toEqual(["Hello", "Hello"]);
+  });
+
+  it("retries once when Edge TTS resolves without creating an output file", async () => {
+    tempDir = mkdtempSync(path.join(tmpdir(), "tts-test-"));
+    const outputPath = path.join(tempDir, "voice.mp3");
+    const calls: string[] = [];
+
+    const deps = createEdgeTTSDeps(async (text: string, filePath: string) => {
+      calls.push(text);
+      if (calls.length === 2) {
+        writeFileSync(filePath, Buffer.from([0xff, 0xfb, 0x90, 0x00]));
+      }
+    });
+
+    await expect(
+      edgeTTS(
+        {
+          text: "Hello",
+          outputPath,
+          config: baseEdgeConfig,
+          timeoutMs: 10000,
+        },
+        deps,
+      ),
+    ).resolves.toBeUndefined();
+    expect(calls).toEqual(["Hello", "Hello"]);
+  });
+
+  it("does not retry provider errors", async () => {
+    tempDir = mkdtempSync(path.join(tmpdir(), "tts-test-"));
+    const outputPath = path.join(tempDir, "voice.mp3");
+    const calls: string[] = [];
+
+    const deps = createEdgeTTSDeps(async (text: string) => {
+      calls.push(text);
+      throw new Error("upstream timeout");
+    });
+
+    await expect(
+      edgeTTS(
+        {
+          text: "Hello",
+          outputPath,
+          config: baseEdgeConfig,
+          timeoutMs: 10000,
+        },
+        deps,
+      ),
+    ).rejects.toThrow("upstream timeout");
+    expect(calls).toEqual(["Hello"]);
+  });
+});

extensions/microsoft/tts.ts

@@ -0,0 +1,138 @@
+// Microsoft plugin module implements tts behavior.
+import { statSync, writeFileSync } from "node:fs";
+import { mkdir } from "node:fs/promises";
+import path from "node:path";
+import { writeExternalFileWithinRoot } from "openclaw/plugin-sdk/security-runtime";
+import { normalizeLowercaseStringOrEmpty } from "openclaw/plugin-sdk/string-coerce-runtime";
+
+type EdgeTTSRuntimeConfig = {
+  voice?: string;
+  lang?: string;
+  outputFormat?: string;
+  saveSubtitles?: boolean;
+  proxy?: string;
+  rate?: string;
+  pitch?: string;
+  volume?: string;
+  timeout?: number;
+};
+
+type EdgeTTSDeps = {
+  EdgeTTS: new (config: EdgeTTSRuntimeConfig) => {
+    ttsPromise: (text: string, outputPath: string) => Promise<unknown>;
+  };
+};
+
+async function loadDefaultEdgeTTSDeps(): Promise<EdgeTTSDeps> {
+  const { EdgeTTS } = await import("node-edge-tts");
+  return { EdgeTTS };
+}
+
+function isMissingOutputFileError(error: unknown): boolean {
+  return (
+    typeof error === "object" &&
+    error !== null &&
+    "code" in error &&
+    (error as { code?: unknown }).code === "ENOENT"
+  );
+}
+
+function readOutputSize(outputPath: string): number {
+  try {
+    return statSync(outputPath).size;
+  } catch (error) {
+    if (isMissingOutputFileError(error)) {
+      return 0;
+    }
+    throw error;
+  }
+}
+
+export function inferEdgeExtension(outputFormat: string): string {
+  const normalized = normalizeLowercaseStringOrEmpty(outputFormat);
+  if (normalized.includes("webm")) {
+    return ".webm";
+  }
+  if (normalized.includes("ogg")) {
+    return ".ogg";
+  }
+  if (normalized.includes("opus")) {
+    return ".opus";
+  }
+  if (normalized.includes("wav") || normalized.includes("riff") || normalized.includes("pcm")) {
+    return ".wav";
+  }
+  return ".mp3";
+}
+
+export async function edgeTTS(
+  params: {
+    text: string;
+    outputPath: string;
+    config: {
+      voice: string;
+      lang: string;
+      outputFormat: string;
+      saveSubtitles: boolean;
+      proxy?: string;
+      rate?: string;
+      pitch?: string;
+      volume?: string;
+      timeoutMs?: number;
+    };
+    timeoutMs: number;
+  },
+  deps?: EdgeTTSDeps,
+): Promise<void> {
+  const { text, outputPath, config, timeoutMs } = params;
+  if (text.trim().length === 0) {
+    throw new Error("Microsoft TTS text cannot be empty");
+  }
+
+  const resolvedDeps = deps ?? (await loadDefaultEdgeTTSDeps());
+  const tts = new resolvedDeps.EdgeTTS({
+    voice: config.voice,
+    lang: config.lang,
+    outputFormat: config.outputFormat,
+    saveSubtitles: config.saveSubtitles,
+    proxy: config.proxy,
+    rate: config.rate,
+    pitch: config.pitch,
+    volume: config.volume,
+    timeout: config.timeoutMs ?? timeoutMs,
+  });
+
+  for (let attempt = 0; attempt < 2; attempt += 1) {
+    const outputSize = await writeEdgeTtsOutput({
+      outputPath,
+      ttsPromise: async (tempPath) => {
+        await tts.ttsPromise(text, tempPath);
+      },
+    });
+    if (outputSize > 0) {
+      return;
+    }
+  }
+  throw new Error("Edge TTS produced empty audio file after retry");
+}
+
+async function writeEdgeTtsOutput(params: {
+  outputPath: string;
+  ttsPromise: (tempPath: string) => Promise<void>;
+}): Promise<number> {
+  const rootDir = path.dirname(params.outputPath);
+  await mkdir(rootDir, { recursive: true });
+  let outputSize = 0;
+  await writeExternalFileWithinRoot({
+    rootDir,
+    path: path.basename(params.outputPath),
+    write: async (tempPath) => {
+      await params.ttsPromise(tempPath);
+      outputSize = readOutputSize(tempPath);
+      if (outputSize === 0) {
+        writeFileSync(tempPath, "");
+      }
+    },
+  });
+  return outputSize;
+}