Increase realtime api websocket frame size to accomodate large model outputs. (#46014)

* Increase max frame size to accomodate model output

Occasionally the last message the model sends of it's out output exceeds 65536 bytes causing the output to cut out. To remedy this the limit can be increased.

* Update sdk/openai/azure-ai-openai-realtime/src/main/java/com/azure/ai/openai/realtime/implementation/websocket/WebSocketSessionNettyImpl.java

Co-authored-by: Jose Alvarez <[email protected]>

* Update WebSocketSessionNettyImpl.java

Added missing import

* Update WebSocketSessionNettyImpl.java

Added trailing space

---------

Co-authored-by: Jose Alvarez <[email protected]>

Author: FrankVanDerVelde

sdk/openai/azure-ai-openai-realtime/src/main/java/com/azure/ai/openai/realtime/implementation/websocket/WebSocketSessionNettyImpl.java

@@ -5,6 +5,7 @@
 
 import com.azure.ai.openai.realtime.models.ConnectFailedException;
 import com.azure.ai.openai.realtime.models.RealtimeClientEvent;
+import com.azure.core.util.Configuration;
 import com.azure.core.util.logging.ClientLogger;
 import io.netty.bootstrap.Bootstrap;
 import io.netty.buffer.PooledByteBufAllocator;
@@ -43,7 +44,7 @@
 
 final class WebSocketSessionNettyImpl implements WebSocketSession {
 
-    private static final int MAX_FRAME_SIZE = 65536;
+    private static final int MAX_FRAME_SIZE = Configuration.getGlobalConfiguration().get("AZURE_MAX_FRAME_SIZE", 65536);
 
     private static final ClientLogger LOGGER = new ClientLogger(WebSocketSessionNettyImpl.class);
     private final MessageEncoder messageEncoder;