fix(speech): bound TTS/STT voice-list and transcription JSON response reads (#96496)

Route success JSON reads through readProviderJsonResponse (16 MiB cap) in
azure-speech, elevenlabs, microsoft, minimax/tts, xai/stt, and
openrouter/media-understanding to prevent OOM from oversized or hostile
endpoint responses. Mirrors the response-limit campaign already applied to
other provider paths.

AI-assisted.

Co-authored-by: Cursor <cursoragent@cursor.com>

extensions/azure-speech/tts.ts

@@ -2,7 +2,10 @@
  * Azure Speech REST helpers. They normalize endpoints, build SSML, list voices,
  * and synthesize speech with response-size and SSRF guards.
  */
-import { assertOkOrThrowProviderError } from "openclaw/plugin-sdk/provider-http";
+import {
+  assertOkOrThrowProviderError,
+  readProviderJsonResponse,
+} from "openclaw/plugin-sdk/provider-http";
 import { readResponseWithLimit } from "openclaw/plugin-sdk/response-limit-runtime";
 import type { SpeechVoiceOption } from "openclaw/plugin-sdk/speech-core";
 import { trimToUndefined } from "openclaw/plugin-sdk/speech-core";
@@ -160,7 +163,10 @@ export async function listAzureSpeechVoices(params: {
 
   try {
     await assertOkOrThrowProviderError(response, "Azure Speech voices API error");
-    const voices = (await response.json()) as AzureSpeechVoiceEntry[];
+    const voices = await readProviderJsonResponse<AzureSpeechVoiceEntry[]>(
+      response,
+      "azure-speech.voices",
+    );
     return Array.isArray(voices)
       ? voices
           .filter((voice) => !isDeprecatedVoice(voice))

extensions/elevenlabs/speech-provider.ts

@@ -1,7 +1,10 @@
 // Elevenlabs provider module implements model/runtime integration.
 import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { parseStrictFiniteNumber, parseStrictInteger } 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,
@@ -367,14 +370,14 @@ async function listElevenLabsVoices(params: {
   });
   try {
     await assertOkOrThrowProviderError(response, "ElevenLabs voices API error");
-    const json = (await response.json()) as {
+    const json = await readProviderJsonResponse<{
       voices?: Array<{
         voice_id?: string;
         name?: string;
         category?: string;
         description?: string;
       }>;
-    };
+    }>(response, "elevenlabs.voices");
     return Array.isArray(json.voices)
       ? json.voices
           .map((voice) => ({

extensions/microsoft/speech-provider.ts

@@ -7,7 +7,10 @@ import {
   generateSecMsGecToken,
 } from "node-edge-tts/dist/drm.js";
 import { isVoiceCompatibleAudio } from "openclaw/plugin-sdk/media-runtime";
-import { assertOkOrThrowProviderError } from "openclaw/plugin-sdk/provider-http";
+import {
+  assertOkOrThrowProviderError,
+  readProviderJsonResponse,
+} from "openclaw/plugin-sdk/provider-http";
 import {
   captureHttpExchange,
   isDebugProxyGlobalFetchPatchInstalled,
@@ -166,7 +169,10 @@ export async function listMicrosoftVoices(): Promise<SpeechVoiceOption[]> {
       });
     }
     await assertOkOrThrowProviderError(response, "Microsoft voices API error");
-    const voices = (await response.json()) as MicrosoftVoiceListEntry[];
+    const voices = await readProviderJsonResponse<MicrosoftVoiceListEntry[]>(
+      response,
+      "microsoft.speech-voices",
+    );
     return Array.isArray(voices)
       ? voices
           .map((voice) => ({

extensions/minimax/tts.ts

@@ -1,6 +1,9 @@
 // Minimax plugin module implements tts behavior.
 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 {
   fetchWithSsrFGuard,
   ssrfPolicyFromHttpBaseUrlAllowedHostname,
@@ -105,10 +108,10 @@ export async function minimaxTTS(params: {
     try {
       await assertOkOrThrowProviderError(response, "MiniMax TTS API error");
 
-      const body = (await response.json()) as {
+      const body = await readProviderJsonResponse<{
         data?: { audio?: string };
         base_resp?: { status_code?: number; status_msg?: string };
-      };
+      }>(response, "minimax.tts");
 
       // Check base_resp for envelope errors (HTTP 200 with non-zero status_code).
       // Other MiniMax providers (image, video, music, web-search) already check this.
@@ -119,9 +122,7 @@ export async function minimaxTTS(params: {
         body.base_resp.status_code !== 0
       ) {
         const msg = body.base_resp.status_msg ?? "unknown error";
-        throw new Error(
-          `MiniMax TTS API error (${body.base_resp.status_code}): ${msg}`,
-        );
+        throw new Error(`MiniMax TTS API error (${body.base_resp.status_code}): ${msg}`);
       }
 
       const hexAudio = body?.data?.audio;

extensions/openrouter/media-understanding-provider.test.ts

@@ -24,6 +24,8 @@ const { assertOkOrThrowHttpErrorMock, postJsonRequestMock, resolveProviderHttpRe
 vi.mock("openclaw/plugin-sdk/provider-http", () => ({
   assertOkOrThrowHttpError: assertOkOrThrowHttpErrorMock,
   postJsonRequest: postJsonRequestMock,
+  // Pass-through: bounded-reader enforcement is tested via bounded-reader unit tests.
+  readProviderJsonResponse: async (response: { json(): Promise<unknown> }) => response.json(),
   requireTranscriptionText: (value: string | undefined, message: string) => {
     const text = value?.trim();
     if (!text) {

extensions/openrouter/media-understanding-provider.ts

@@ -10,6 +10,7 @@ import {
 import {
   assertOkOrThrowHttpError,
   postJsonRequest,
+  readProviderJsonResponse,
   requireTranscriptionText,
   resolveProviderHttpRequestConfig,
 } from "openclaw/plugin-sdk/provider-http";
@@ -148,7 +149,10 @@ export async function transcribeOpenRouterAudio(
 
   try {
     await assertOkOrThrowHttpError(response, "OpenRouter audio transcription failed");
-    const payload = (await response.json()) as OpenRouterSttResponse;
+    const payload = await readProviderJsonResponse<OpenRouterSttResponse>(
+      response,
+      "openrouter.stt",
+    );
     return {
       text: requireTranscriptionText(
         payload.text,

extensions/xai/stt.ts

@@ -8,8 +8,9 @@ import {
   assertOkOrThrowHttpError,
   buildAudioTranscriptionFormData,
   postTranscriptionRequest,
-  resolveProviderHttpRequestConfig,
+  readProviderJsonResponse,
   requireTranscriptionText,
+  resolveProviderHttpRequestConfig,
 } from "openclaw/plugin-sdk/provider-http";
 import { normalizeOptionalString } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { XAI_BASE_URL } from "./model-definitions.js";
@@ -68,7 +69,7 @@ export async function transcribeXaiAudio(
 
   try {
     await assertOkOrThrowHttpError(response, "xAI audio transcription failed");
-    const payload = (await response.json()) as XaiSttResponse;
+    const payload = await readProviderJsonResponse<XaiSttResponse>(response, "xai.stt");
     return {
       text: requireTranscriptionText(payload.text, "xAI transcription response missing text"),
       ...(model ? { model } : {}),