重构主控编排与运行时预热链路，统一工作区提示词/专家调度协议并补齐 wiki 记忆注入与写回闭环。

同时收敛启动与运维脚本默认行为（含 wiki worker）、更新 Admin 可观测性与相关测试，降低首轮时延并提高运行稳定性。

Made-with: Cursor

openclaw/extensions/speech-core/api.ts

@@ -0,0 +1 @@
+export * from "openclaw/plugin-sdk/speech-core";

openclaw/extensions/speech-core/package.json

@@ -0,0 +1,10 @@
+{
+  "name": "@openclaw/speech-core",
+  "version": "2026.4.20",
+  "private": true,
+  "description": "OpenClaw speech runtime package",
+  "type": "module",
+  "devDependencies": {
+    "@openclaw/plugin-sdk": "workspace:*"
+  }
+}

openclaw/extensions/speech-core/runtime-api.ts

@@ -0,0 +1,34 @@
+export {
+  buildTtsSystemPromptHint,
+  getLastTtsAttempt,
+  getResolvedSpeechProviderConfig,
+  getTtsMaxLength,
+  getTtsProvider,
+  isSummarizationEnabled,
+  isTtsEnabled,
+  isTtsProviderConfigured,
+  listSpeechVoices,
+  maybeApplyTtsToPayload,
+  resolveExplicitTtsOverrides,
+  resolveTtsAutoMode,
+  resolveTtsConfig,
+  resolveTtsPrefsPath,
+  resolveTtsProviderOrder,
+  setLastTtsAttempt,
+  setSummarizationEnabled,
+  setTtsAutoMode,
+  setTtsEnabled,
+  setTtsMaxLength,
+  setTtsProvider,
+  synthesizeSpeech,
+  textToSpeech,
+  textToSpeechTelephony,
+  _test,
+  type ResolvedTtsConfig,
+  type ResolvedTtsModelOverrides,
+  type TtsDirectiveOverrides,
+  type TtsDirectiveParseResult,
+  type TtsResult,
+  type TtsSynthesisResult,
+  type TtsTelephonyResult,
+} from "./src/tts.js";

openclaw/extensions/speech-core/src/tts.test.ts

@@ -0,0 +1,140 @@
+import { rmSync } from "node:fs";
+import path from "node:path";
+import type { OpenClawConfig } from "openclaw/plugin-sdk/config-runtime";
+import type { ReplyPayload } from "openclaw/plugin-sdk/reply-payload";
+import type { SpeechProviderPlugin, SpeechSynthesisRequest } from "openclaw/plugin-sdk/speech-core";
+import { afterEach, describe, expect, it, vi } from "vitest";
+
+type MockSpeechSynthesisResult = Awaited<ReturnType<SpeechProviderPlugin["synthesize"]>>;
+
+const synthesizeMock = vi.hoisted(() =>
+  vi.fn(
+    async (request: SpeechSynthesisRequest): Promise<MockSpeechSynthesisResult> => ({
+      audioBuffer: Buffer.from("voice"),
+      fileExtension: ".ogg",
+      outputFormat: "ogg",
+      voiceCompatible: request.target === "voice-note",
+    }),
+  ),
+);
+
+const listSpeechProvidersMock = vi.hoisted(() => vi.fn());
+const getSpeechProviderMock = vi.hoisted(() => vi.fn());
+
+vi.mock("openclaw/plugin-sdk/channel-targets", () => ({
+  normalizeChannelId: (channel: string | undefined) => channel?.trim().toLowerCase() ?? null,
+}));
+
+vi.mock("../api.js", async () => {
+  const actual = await vi.importActual<typeof import("../api.js")>("../api.js");
+  const mockProvider: SpeechProviderPlugin = {
+    id: "mock",
+    label: "Mock",
+    autoSelectOrder: 1,
+    isConfigured: () => true,
+    synthesize: synthesizeMock,
+  };
+  listSpeechProvidersMock.mockImplementation(() => [mockProvider]);
+  getSpeechProviderMock.mockImplementation((providerId: string) =>
+    providerId === "mock" ? mockProvider : null,
+  );
+  return {
+    ...actual,
+    canonicalizeSpeechProviderId: (providerId: string | undefined) =>
+      providerId?.trim().toLowerCase() || undefined,
+    normalizeSpeechProviderId: (providerId: string | undefined) =>
+      providerId?.trim().toLowerCase() || undefined,
+    getSpeechProvider: getSpeechProviderMock,
+    listSpeechProviders: listSpeechProvidersMock,
+    scheduleCleanup: vi.fn(),
+  };
+});
+
+const { _test, maybeApplyTtsToPayload } = await import("./tts.js");
+
+const nativeVoiceNoteChannels = ["discord", "feishu", "matrix", "telegram", "whatsapp"] as const;
+
+function createTtsConfig(prefsName: string): OpenClawConfig {
+  return {
+    messages: {
+      tts: {
+        enabled: true,
+        provider: "mock",
+        prefsPath: `/tmp/${prefsName}.json`,
+      },
+    },
+  };
+}
+
+describe("speech-core native voice-note routing", () => {
+  afterEach(() => {
+    synthesizeMock.mockClear();
+  });
+
+  it("keeps native voice-note channel support centralized", () => {
+    for (const channel of nativeVoiceNoteChannels) {
+      expect(_test.supportsNativeVoiceNoteTts(channel)).toBe(true);
+      expect(_test.supportsNativeVoiceNoteTts(channel.toUpperCase())).toBe(true);
+    }
+    expect(_test.supportsNativeVoiceNoteTts("slack")).toBe(false);
+    expect(_test.supportsNativeVoiceNoteTts(undefined)).toBe(false);
+  });
+
+  it("marks Discord auto TTS replies as native voice messages", async () => {
+    const cfg = createTtsConfig("openclaw-speech-core-tts-test");
+    const payload: ReplyPayload = {
+      text: "This Discord reply should be delivered as a native voice note.",
+    };
+
+    let mediaDir: string | undefined;
+    try {
+      const result = await maybeApplyTtsToPayload({
+        payload,
+        cfg,
+        channel: "discord",
+        kind: "final",
+      });
+
+      expect(synthesizeMock).toHaveBeenCalledWith(
+        expect.objectContaining({ target: "voice-note" }),
+      );
+      expect(result.audioAsVoice).toBe(true);
+      expect(result.mediaUrl).toMatch(/voice-\d+\.ogg$/);
+
+      mediaDir = result.mediaUrl ? path.dirname(result.mediaUrl) : undefined;
+    } finally {
+      if (mediaDir) {
+        rmSync(mediaDir, { recursive: true, force: true });
+      }
+    }
+  });
+
+  it("keeps non-native voice-note channels as regular audio files", async () => {
+    const cfg = createTtsConfig("openclaw-speech-core-tts-slack-test");
+    const payload: ReplyPayload = {
+      text: "Slack replies should be delivered as regular audio attachments.",
+    };
+
+    let mediaDir: string | undefined;
+    try {
+      const result = await maybeApplyTtsToPayload({
+        payload,
+        cfg,
+        channel: "slack",
+        kind: "final",
+      });
+
+      expect(synthesizeMock).toHaveBeenCalledWith(
+        expect.objectContaining({ target: "audio-file" }),
+      );
+      expect(result.audioAsVoice).toBeUndefined();
+      expect(result.mediaUrl).toMatch(/voice-\d+\.ogg$/);
+
+      mediaDir = result.mediaUrl ? path.dirname(result.mediaUrl) : undefined;
+    } finally {
+      if (mediaDir) {
+        rmSync(mediaDir, { recursive: true, force: true });
+      }
+    }
+  });
+});

openclaw/extensions/speech-core/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"
+  ]
+}