run-llama/llama_index · error · ValueError

Invalid

Error message

Invalid

What it means

Poorly-named ValueError('Invalid') raised in the sync predict_and_call() flow: after executing the model's tool calls, if more than one tool output was produced while allow_parallel_tool_calls=False, the code cannot fold multiple results into a single AgentChatResponse and raises. It signals the model issued parallel tool calls even though they were disallowed.

Source

Thrown at llama-index-core/llama_index/core/llms/function_calling.py:258

            call_tool_with_selection(tool_call, tools, verbose=verbose)
            for tool_call in tool_calls
        ]
        tool_outputs_with_error = [
            tool_output for tool_output in tool_outputs if tool_output.is_error
        ]
        if error_on_tool_error and len(tool_outputs_with_error) > 0:
            error_text = "\n\n".join(
                [tool_output.content for tool_output in tool_outputs]
            )
            raise ValueError(error_text)
        elif allow_parallel_tool_calls:
            output_text = "\n\n".join(
                [tool_output.content for tool_output in tool_outputs]
            )
            return AgentChatResponse(response=output_text, sources=tool_outputs)
        else:
            if len(tool_outputs) > 1:
                raise ValueError("Invalid")
            elif len(tool_outputs) == 0:
                return AgentChatResponse(
                    response=response.message.content or "", sources=tool_outputs
                )

            return AgentChatResponse(
                response=tool_outputs[0].content, sources=tool_outputs
            )

    async def apredict_and_call(
        self,
        tools: Sequence["BaseTool"],
        user_msg: Optional[Union[str, ChatMessage]] = None,
        chat_history: Optional[List[ChatMessage]] = None,
        verbose: bool = False,
        allow_parallel_tool_calls: bool = False,
        error_on_no_tool_call: bool = True,
        error_on_tool_error: bool = False,

View on GitHub (pinned to afd0fef371)

Solutions

1. Pass allow_parallel_tool_calls=True if your agent can handle multiple results.
2. Disable parallel tool calls at the provider level (e.g. OpenAI client with parallel_tool_calls=False) so the model returns one call per turn.
3. Catch ValueError here and retry with a stronger instruction, or use a higher-level agent runner that normalizes multi-call responses.

Example fix

# before
resp = llm.predict_and_call(tools, user_msg='do both tasks')  # model returns 2 tool calls -> ValueError('Invalid')

# after
resp = llm.predict_and_call(tools, user_msg='do both tasks', allow_parallel_tool_calls=True)

try:
    resp = llm.predict_and_call(tools, user_msg, allow_parallel_tool_calls=True)
except ValueError:
    resp = llm.predict_and_call(tools, user_msg, allow_parallel_tool_calls=False)

try:
    resp = llm.predict_and_call(tools, user_msg, allow_parallel_tool_calls=False)
except ValueError as e:
    if str(e) == 'Invalid':
        resp = llm.predict_and_call(tools, user_msg, allow_parallel_tool_calls=True)
    else:
        raise

When it happens

Trigger: llm.predict_and_call(tools, user_msg, allow_parallel_tool_calls=False) where the underlying model returns multiple tool_calls in one response — possible with models like GPT-4o that like parallel calls, or when the provider ignored the parallel-calls flag. get_tool_calls_from_response returns several selections, all execute, and len(tool_outputs) > 1 hits the branch.

Common situations: Agents built on predict_and_call with default flags against models that aggressively batch tool calls; provider config that enables parallel function calling at the API level (e.g. OpenAI parallel_tool_calls=True) contradicting the llama-index flag.

Related errors

• get_tool_calls_from_response is not supported by default.
• Max iterations of {max_iterations} reached! Either something
• No tool calls found, cannot aggregate results.
• Tool {tool.metadata.name} requires context. CodeActAgent onl
• Tool {tool.metadata.name} is not a FunctionTool. CodeActAgen

AI-assisted analysis of run-llama/llama_index@afd0fef371 (2026-08-15). Data as JSON: /api/errors/e8d306447ca4c612. Report an issue: GitHub.