ai-agents-pydantic-validation--when-using-langserve-with-an-inputchat-model-that--5de03e16

THEN Replace the typed Union list with a plain List field in your InputChat Pydantic model (e.g., chat_history: List = Field(...)). This bypasses the overly strict pydantic v2 union validation that expects a 'tool_call_id' key from non-tool messages. The underlying fix in langchain_core's ToolMessage.coerce_args is to use values.get('tool_call_id') instead of values['tool_call_id'].

class InputChat(BaseModel):
    chat_history: List = Field(
        ...,
        description="The chat messages representing the current conversation."
    )
    input: str = Field(...)