fix(openai): bound realtime voice websocket payload at 16 MiB (#99450)

sunlit-deng · steipete · web-flow · commit 2525ea41fd09 · 2026-07-06T00:12:27.000+01:00
* fix(openai): bound realtime voice websocket payload at 16 MiB

* test(openai): cover realtime voice payload cap

---------

Co-authored-by: Peter Steinberger &lt;steipete@gmail.com&gt;
diff --git a/extensions/openai/realtime-voice-provider.test.ts b/extensions/openai/realtime-voice-provider.test.ts
@@ -333,13 +333,16 @@ describe("buildOpenAIRealtimeVoiceProvider", () => {
     bridge.close();
 
     const socket = FakeWebSocket.instances[0];
-    const options = socket?.args[1] as { headers?: Record<string, string> } | undefined;
+    const options = socket?.args[1] as
+      | { headers?: Record<string, string>; maxPayload?: number }
+      | undefined;
     expectRecordFields(options?.headers, "websocket headers", {
       originator: "openclaw",
       version: "2026.3.22",
       "User-Agent": "openclaw/2026.3.22",
     });
     expect(options?.headers).not.toHaveProperty("OpenAI-Beta");
+    expect(options?.maxPayload).toBe(16 * 1024 * 1024);
   });
 
   it("requires a Platform API key for native realtime websocket bridges", async () => {
diff --git a/extensions/openai/realtime-voice-provider.ts b/extensions/openai/realtime-voice-provider.ts
@@ -93,6 +93,7 @@ const OPENAI_REALTIME_ACTIVE_RESPONSE_ERROR_PREFIX =
 const OPENAI_REALTIME_NO_ACTIVE_RESPONSE_CANCEL_ERROR =
   "Cancellation failed: no active response found";
 const OPENAI_REALTIME_MAX_SESSION_DURATION_FRAGMENT = "maximum duration";
+const OPENAI_VOICE_WS_MAX_PAYLOAD_BYTES = 16 * 1024 * 1024;
 const OPENAI_REALTIME_DEFAULT_MIN_BARGE_IN_AUDIO_END_MS = 250;
 // Realtime validates this character set but accepts names beyond the 64-character
 // cap used by other OpenAI tool surfaces.
@@ -616,6 +617,7 @@ class OpenAIRealtimeVoiceBridge implements RealtimeVoiceBridge {
         const proxyAgent = createDebugProxyWebSocketAgent(debugProxy);
         const ws = new WebSocket(connection.url, {
           headers: connection.headers,
+          maxPayload: OPENAI_VOICE_WS_MAX_PAYLOAD_BYTES,
           ...(proxyAgent ? { agent: proxyAgent } : {}),
         });
         this.ws = ws;
