BerriAI/litellm · error · ValueError

Error updating current item chunks: {e}, got transformed_mes

Error message

Error updating current item chunks: {e}, got transformed_message: {transformed_message}

What it means

Raised in update_current_item_chunks, the item-level counterpart of the delta-chunk accumulator. It reads transformed_message['type'] looking for 'response.output_item.done'; any exception (non-dict input, missing 'type') is caught and re-raised as ValueError embedding the transformed_message. Guards internal state consistency while assembling completed output items.

Source

Thrown at litellm/llms/gemini/realtime/transformation.py:964

            if isinstance(transformed_message, list):
                current_item_chunks = []
                any_item_chunk = False
                for event in transformed_message:
                    if event["type"] == "response.output_item.done":
                        current_item_chunks.append(cast(OpenAIRealtimeOutputItemDone, event))
                        any_item_chunk = True
                if not any_item_chunk:
                    current_item_chunks = None
            else:
                if transformed_message["type"] == "response.output_item.done":
                    if current_item_chunks is None:
                        current_item_chunks = []
                    current_item_chunks.append(cast(OpenAIRealtimeOutputItemDone, transformed_message))
                else:
                    current_item_chunks = None
            return current_item_chunks
        except Exception as e:
            raise ValueError(f"Error updating current item chunks: {e}, got transformed_message: {transformed_message}")

    def transform_response_done_event(
        self,
        message: BidiGenerateContentServerMessage,
        current_response_id: str | None,
        current_conversation_id: str | None,
        output_items: list[OpenAIRealtimeOutputItemDone] | None,
        session_configuration_request: str | None = None,
    ) -> OpenAIRealtimeDoneEvent:
        if current_conversation_id is None:
            current_conversation_id = f"conv_{uuid.uuid4()}"
        if current_response_id is None:
            current_response_id = f"resp_{uuid.uuid4()}"

        if session_configuration_request:
            session_configuration_request_dict: BidiGenerateContentSetup = json.loads(
                session_configuration_request
            ).get("setup", {})

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Feed only transformer-output events (OpenAIRealtimeEvents) into this helper.
2. Add a shape check (dict with 'type') before calling.
3. Upgrade litellm when the error originates inside the library loop.

Example fix

# before
items = config.update_current_item_chunks(transformed_message=frame, current_item_chunks=items)

# after
if not (isinstance(frame, dict) and 'type' in frame):
    raise TypeError('expected transformed OpenAI realtime event dict')
items = config.update_current_item_chunks(transformed_message=frame, current_item_chunks=items)

def is_transformed_event(msg) -> bool:
    return isinstance(msg, dict) and "type" in msg

def is_transformed_event(msg: object) -> bool:
    return isinstance(msg, dict) and isinstance(msg.get("type"), str)

try:
    items = config.update_current_item_chunks(transformed_message=msg, current_item_chunks=items)
except ValueError as e:
    if "Error updating current item chunks" in str(e):
        logger.error("Item state update failed on: %r", msg)
        items = None
    else:
        raise

When it happens

Trigger: Passing a transformed_message that is not a dict with a 'type' key (raw string, Gemini-native frame, None) into update_current_item_chunks; list inputs whose elements are malformed.

Common situations: Direct use of the realtime transformer's state helpers in custom bridges; version mismatches between the transformer and hand-rolled event producers; test fixtures with simplified event dicts.

Related errors

• Error updating current delta chunks: {e}, got transformed_me
• Unexpected delta type: {delta_type}
• Unknown openai event: {openai_event}
• Unexpected responses stream payload
• api_key is required for Gemini API calls

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/f05e4cf287afdf72. Report an issue: GitHub.