From a8b592b634652926f6c7f4a1c38741cf5fafadc8 Mon Sep 17 00:00:00 2001
From: Cursor Agent <cursoragent@cursor.com>
Date: Sat, 23 May 2026 20:29:56 +0000
Subject: [PATCH] fix(gemini realtime): event_id, item copy, and dict guard for
 tool-call events

- Emit event_id on response.output_item.added for tool calls so spec-compliant
  OpenAI Realtime SDK clients can index/deduplicate the event like every other
  server-sent event in the sequence.
- Pass a shallow copy of function_call_item to response.output_item.done and
  conversation.item.created so downstream handlers (e.g. the beta-protocol
  translator) that mutate the item dict don't corrupt sibling events sharing
  the same reference.
- Guard map_openai_event against non-dict values (e.g. Gemini's
  'setupComplete: true' boolean payload) so the WebSocket session doesn't die
  with an AttributeError on the unguarded .get() call.

Add NotRequired event_id field on OpenAIRealtimeStreamResponseOutputItemAdded
to keep existing call-sites that don't set event_id compatible.

Co-authored-by: Yassin Kortam <yassin@berri.ai>
---
 .../llms/gemini/realtime/transformation.py    | 20 +++++++++++++------
 litellm/types/llms/openai.py                  | 10 +++++++++-
 2 files changed, 23 insertions(+), 7 deletions(-)

diff --git a/litellm/llms/gemini/realtime/transformation.py b/litellm/llms/gemini/realtime/transformation.py
index 2346d3b7a73..0c0baaeaf8f 100644
--- a/litellm/llms/gemini/realtime/transformation.py
+++ b/litellm/llms/gemini/realtime/transformation.py
@@ -1138,11 +1138,15 @@ class GeminiRealtimeConfig(BaseRealtimeConfig):
     def map_openai_event(
         self,
         key: str,
-        value: dict,
+        value: Any,
         current_delta_type: Optional[ALL_DELTA_TYPES],
     ) -> Union[OpenAIRealtimeEventTypes, ResponsesAPIStreamEvents]:
-        model_turn_event = value.get("modelTurn")
-        generation_complete_event = value.get("generationComplete")
+        if isinstance(value, dict):
+            model_turn_event = value.get("modelTurn")
+            generation_complete_event = value.get("generationComplete")
+        else:
+            model_turn_event = None
+            generation_complete_event = None
         openai_event: Optional[
             Union[OpenAIRealtimeEventTypes, ResponsesAPIStreamEvents]
         ] = None
@@ -1399,6 +1403,7 @@ class GeminiRealtimeConfig(BaseRealtimeConfig):
                     returned_message.append(
                         OpenAIRealtimeStreamResponseOutputItemAdded(
                             type="response.output_item.added",
+                            event_id=f"event_{uuid.uuid4()}",
                             response_id=current_response_id,
                             output_index=idx,
                             item={
@@ -1431,14 +1436,17 @@ class GeminiRealtimeConfig(BaseRealtimeConfig):
                     )
                     # response.function_call_arguments.done
                     returned_message.append(tool_call)
-                    # response.output_item.done
+                    # response.output_item.done — pass a fresh copy so
+                    # downstream handlers that mutate the item dict (e.g. the
+                    # beta-protocol translator) don't corrupt the references
+                    # used by sibling events sharing the same function_call_item.
                     returned_message.append(
                         OpenAIRealtimeOutputItemDone(
                             type="response.output_item.done",
                             event_id=f"event_{uuid.uuid4()}",
                             response_id=current_response_id,
                             output_index=idx,
-                            item=function_call_item,
+                            item={**function_call_item},
                         )
                     )
                     # conversation.item.created
@@ -1446,7 +1454,7 @@ class GeminiRealtimeConfig(BaseRealtimeConfig):
                         OpenAIRealtimeConversationItemCreated(
                             type="conversation.item.created",
                             event_id=f"event_{uuid.uuid4()}",
-                            item=function_call_item,
+                            item={**function_call_item},
                         )
                     )
 
diff --git a/litellm/types/llms/openai.py b/litellm/types/llms/openai.py
index 1ca55534441..14114b22f39 100644
--- a/litellm/types/llms/openai.py
+++ b/litellm/types/llms/openai.py
@@ -79,7 +79,14 @@ from pydantic import (
     field_serializer,
     field_validator,
 )
-from typing_extensions import Annotated, Dict, Required, TypedDict, override
+from typing_extensions import (
+    Annotated,
+    Dict,
+    NotRequired,
+    Required,
+    TypedDict,
+    override,
+)
 
 from litellm.types.llms.base import BaseLiteLLMOpenAIResponseObject
 from litellm.types.responses.main import (
@@ -1935,6 +1942,7 @@ class OpenAIRealtimeStreamResponseOutputItemAdded(TypedDict):
     response_id: str
     output_index: int
     item: OpenAIRealtimeStreamResponseOutputItem
+    event_id: NotRequired[str]
 
 
 class OpenAIRealtimeStreamResponseBaseObject(TypedDict):