fix(speech): bound TTS response reads (#96874)

(cherry picked from commit 2f851ecfe9)

extensions/volcengine/tts.test.ts

@@ -7,6 +7,8 @@ const { fetchWithSsrFGuardMock } = vi.hoisted(() => ({
   fetchWithSsrFGuardMock: vi.fn(),
 }));
 
+const PROVIDER_RESPONSE_MAX_BYTES = 16 * 1024 * 1024;
+
 vi.mock("openclaw/plugin-sdk/ssrf-runtime", () => ({
   fetchWithSsrFGuard: fetchWithSsrFGuardMock,
 }));
@@ -45,6 +47,18 @@ function clearTtsEnv() {
   delete process.env.VOLCENGINE_TTS_TOKEN;
 }
 
+function makeOversizedStreamResponse(): Response {
+  return new Response(
+    new ReadableStream<Uint8Array>({
+      start(controller) {
+        controller.enqueue(new Uint8Array(PROVIDER_RESPONSE_MAX_BYTES));
+        controller.enqueue(new Uint8Array(1));
+        controller.close();
+      },
+    }),
+  );
+}
+
 function restoreOptionalEnv(key: string, value: string | undefined) {
   if (value === undefined) {
     delete process.env[key];
@@ -301,6 +315,23 @@ describe("volcengineTTS", () => {
     expect(release).toHaveBeenCalledTimes(1);
   });
 
+  it("bounds Seed Speech success response reads", async () => {
+    const release = vi.fn();
+    fetchWithSsrFGuardMock.mockResolvedValue({
+      response: makeOversizedStreamResponse(),
+      release,
+    });
+
+    await expect(
+      volcengineTTS({
+        text: "hello",
+        apiKey: "secret-api-key",
+        timeoutMs: 1000,
+      }),
+    ).rejects.toThrow("BytePlus Seed Speech TTS response exceeds 16777216 bytes");
+    expect(release).toHaveBeenCalledTimes(1);
+  });
+
   it("reports provider errors without exposing credentials", async () => {
     const release = vi.fn();
     fetchWithSsrFGuardMock.mockResolvedValue({
@@ -327,4 +358,22 @@ describe("volcengineTTS", () => {
     expect((error as Error).message).not.toContain("secret-token");
     expect(release).toHaveBeenCalledTimes(1);
   });
+
+  it("bounds legacy Volcengine success response reads", async () => {
+    const release = vi.fn();
+    fetchWithSsrFGuardMock.mockResolvedValue({
+      response: makeOversizedStreamResponse(),
+      release,
+    });
+
+    await expect(
+      volcengineTTS({
+        text: "hello",
+        appId: "app-id",
+        token: "secret-token",
+        timeoutMs: 1000,
+      }),
+    ).rejects.toThrow("Volcengine TTS response exceeds 16777216 bytes");
+    expect(release).toHaveBeenCalledTimes(1);
+  });
 });

extensions/volcengine/tts.ts

@@ -1,5 +1,6 @@
 // Volcengine plugin module implements tts behavior.
 import * as crypto from "node:crypto";
+import { readResponseWithLimit } from "openclaw/plugin-sdk/response-limit-runtime";
 import { fetchWithSsrFGuard } from "openclaw/plugin-sdk/ssrf-runtime";
 
 export type VolcengineTtsEncoding = "ogg_opus" | "mp3" | "pcm" | "wav";
@@ -27,6 +28,7 @@ const DEFAULT_LEGACY_VOICE = "zh_female_xiaohe_uranus_bigtts";
 const DEFAULT_CLUSTER = "volcano_tts";
 const DEFAULT_SEED_TTS_RESOURCE_ID = "seed-tts-1.0";
 const DEFAULT_SEED_TTS_APP_KEY = "aGjiRDfUWi";
+const VOLCENGINE_TTS_RESPONSE_MAX_BYTES = 16 * 1024 * 1024;
 const BYTEPLUS_SEED_TTS_URL =
   "https://voice.ap-southeast-1.bytepluses.com/api/v3/tts/unidirectional";
 const VOLCENGINE_LEGACY_TTS_URL = "https://openspeech.bytedance.com/api/v1/tts";
@@ -158,7 +160,13 @@ async function seedSpeechTTS(params: VolcengineTTSParams & { apiKey: string }):
   });
 
   try {
-    const frames = parseSeedTtsFrames(await response.text());
+    const responseText = new TextDecoder().decode(
+      await readResponseWithLimit(response, VOLCENGINE_TTS_RESPONSE_MAX_BYTES, {
+        onOverflow: ({ maxBytes }) =>
+          new Error(`BytePlus Seed Speech TTS response exceeds ${maxBytes} bytes`),
+      }),
+    );
+    const frames = parseSeedTtsFrames(responseText);
     const chunks: Buffer[] = [];
     for (const frame of frames) {
       if (frame.code === 0) {
@@ -240,7 +248,13 @@ async function legacyVolcengineTTS(
   });
 
   try {
-    const body = parseLegacyTtsResponse(await response.text());
+    const responseText = new TextDecoder().decode(
+      await readResponseWithLimit(response, VOLCENGINE_TTS_RESPONSE_MAX_BYTES, {
+        onOverflow: ({ maxBytes }) =>
+          new Error(`Volcengine TTS response exceeds ${maxBytes} bytes`),
+      }),
+    );
+    const body = parseLegacyTtsResponse(responseText);
     if (!response.ok || body.code !== 3000 || !body.data) {
       throw new Error(
         `Volcengine TTS error ${body.code ?? response.status}: ${body.message ?? "unknown"}`,

extensions/xiaomi/speech-provider.test.ts

@@ -4,12 +4,30 @@ import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 
 const transcodeAudioBufferToOpusMock = vi.hoisted(() => vi.fn());
 
+const PROVIDER_RESPONSE_MAX_BYTES = 16 * 1024 * 1024;
+
 vi.mock("openclaw/plugin-sdk/media-runtime", () => ({
   transcodeAudioBufferToOpus: transcodeAudioBufferToOpusMock,
 }));
 
 import { buildXiaomiSpeechProvider } from "./speech-provider.js";
 
+function makeOversizedStreamResponse(): Response {
+  return new Response(
+    new ReadableStream<Uint8Array>({
+      start(controller) {
+        controller.enqueue(new Uint8Array(PROVIDER_RESPONSE_MAX_BYTES));
+        controller.enqueue(new Uint8Array(1));
+        controller.close();
+      },
+    }),
+    {
+      status: 200,
+      headers: { "Content-Type": "application/json" },
+    },
+  );
+}
+
 describe("buildXiaomiSpeechProvider", () => {
   const provider = buildXiaomiSpeechProvider();
 
@@ -410,5 +428,19 @@ describe("buildXiaomiSpeechProvider", () => {
         }),
       ).rejects.toThrow("Xiaomi TTS API returned no audio data");
     });
+
+    it("bounds oversized Xiaomi TTS success response reads", async () => {
+      vi.mocked(globalThis.fetch).mockResolvedValueOnce(makeOversizedStreamResponse());
+
+      await expect(
+        provider.synthesize({
+          text: "Test",
+          cfg: {} as never,
+          providerConfig: { apiKey: "sk-test" },
+          target: "audio-file",
+          timeoutMs: 30000,
+        }),
+      ).rejects.toThrow("Xiaomi TTS API: JSON response exceeds 16777216 bytes");
+    });
   });
 });

extensions/xiaomi/speech-provider.ts

@@ -1,7 +1,10 @@
 // Xiaomi provider module implements model/runtime integration.
 import { transcodeAudioBufferToOpus } from "openclaw/plugin-sdk/media-runtime";
 import { resolveTimerTimeoutMs } from "openclaw/plugin-sdk/number-runtime";
-import { assertOkOrThrowProviderError } from "openclaw/plugin-sdk/provider-http";
+import {
+  assertOkOrThrowProviderError,
+  readProviderJsonResponse,
+} from "openclaw/plugin-sdk/provider-http";
 import { normalizeResolvedSecretInputString } from "openclaw/plugin-sdk/secret-input";
 import type {
   SpeechDirectiveTokenParseContext,
@@ -269,7 +272,8 @@ async function xiaomiTTS(params: {
     });
     try {
       await assertOkOrThrowProviderError(response, "Xiaomi TTS API error");
-      return decodeXiaomiAudioData(await response.json());
+      const body = await readProviderJsonResponse<unknown>(response, "Xiaomi TTS API");
+      return decodeXiaomiAudioData(body);
     } finally {
       await release();
     }