BerriAI/litellm · error · Exception

function_call missing. Received tool call with 'type': 'func

Error message

function_call missing. Received tool call with 'type': 'function'. No function call in argument - {tool}

What it means

While converting an OpenAI tool_calls list to Gemini format, a tool entry had type 'function' but its `function` payload (name/arguments) was missing or unusable, so the helper returned None. LiteLLM deliberately raises instead of silently dropping the malformed tool call. The whole message is included so you can see which entry is bad.

Source

Thrown at litellm/litellm_core_utils/prompt_templates/factory.py:1330

            VertexGeminiConfig,
        )

        if tool_calls is not None:
            for idx, tool in enumerate(tool_calls):
                if "function" in tool:
                    gemini_function_call: VertexFunctionCall | None = _gemini_tool_call_invoke_helper(
                        function_call_params=tool["function"],
                        tool_call_id=(tool.get("id") if forward_function_call_id else None),
                    )
                    if gemini_function_call is not None:
                        part_dict: VertexPartType = {"function_call": gemini_function_call}
                        thought_signature = _get_thought_signature_from_tool(dict(tool), model=model)
                        if thought_signature:
                            part_dict["thoughtSignature"] = thought_signature

                        _parts_list.append(part_dict)
                    else:  # don't silently drop params. Make it clear to user what's happening.
                        raise Exception(
                            f"function_call missing. Received tool call with 'type': 'function'. No function call in argument - {tool}"
                        )
        elif function_call is not None:
            gemini_function_call = _gemini_tool_call_invoke_helper(function_call_params=function_call)
            if gemini_function_call is not None:
                part_dict_function: Final[VertexPartType] = {"function_call": gemini_function_call}

                # Extract thought signature from function_call's provider_specific_fields
                thought_signature = None
                provider_fields: Final = (
                    function_call.get("provider_specific_fields") if isinstance(function_call, dict) else {}
                )
                if isinstance(provider_fields, dict):
                    thought_signature = provider_fields.get("thought_signature")

                # If no signature found and model is gemini-3, use dummy signature
                if not thought_signature and model and VertexGeminiConfig._is_gemini_3_or_newer(model):
                    thought_signature = _get_dummy_thought_signature()

View on GitHub (pinned to 6c2dcb801b)

Solutions

Inspect the offending entry in the error message and fix its 'function' dict to include a valid 'name' (and 'arguments')
If you build tool_calls yourself, always include {"type":"function","function":{"name":..., "arguments":"{}"}}
Sanitize stored histories: drop or repair tool_calls entries whose function/name is falsy before sending to Gemini

Example fix

# before
history.append({
    "role": "assistant",
    "tool_calls": [{"id": "call_1", "type": "function", "function": {}}],
})
litellm.completion(model="gemini/gemini-2.0-flash", messages=history, tools=tools)

# after
history.append({
    "role": "assistant",
    "tool_calls": [{"id": "call_1", "type": "function",
                     "function": {"name": "get_weather", "arguments": "{\"city\": \"SF\"}"}}],
})

Defensive patterns

Strategy: validation

Validate before calling

def tool_calls_are_well_formed(tool_calls) -> bool:
    for tc in tool_calls or []:
        fn = tc.get("function") if isinstance(tc, dict) else None
        if not isinstance(fn, dict) or not fn.get("name"):
            return False
    return True

for m in messages:
    if m.get("role") == "assistant" and not tool_calls_are_well_formed(m.get("tool_calls")):
        raise ValueError(f"malformed tool_calls in assistant message: {m}")

Type guard

from typing import Any

def is_valid_tool_call(tc: Any) -> bool:
    return (
        isinstance(tc, dict)
        and tc.get("type") == "function"
        and isinstance(tc.get("function"), dict)
        and bool(tc["function"].get("name"))
    )

Prevention

Always include function.name and arguments when constructing assistant tool_calls
Validate persisted histories after JSON round-trips
Never replay tool calls whose function dict is empty

When it happens

Trigger: An assistant message with tool_calls=[{"id":..., "type":"function", "function": None}] or function lacking a 'name'; tool calls reconstructed from logs/DB where the function dict was dropped; calling Vertex/Gemini models after resuming an agent conversation with corrupted tool-call history.

Common situations: Replaying persisted conversations where JSON round-trips dropped nested function objects; hand-crafted assistant tool_calls in tests; upstream provider returned a degenerate tool call that was stored verbatim.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/2afb5e59e2485cd9. Report an issue: GitHub.