run-llama/llama_index · error · ValueError

No prompt provided in positional or keyword arguments

Error message

No prompt provided in positional or keyword arguments

What it means

Raised by the prompt-extraction helper inside the LLM callback-decorator wrapper (llms/callbacks.py) when neither a positional argument nor a 'prompt' keyword argument is present on the wrapped predict/complete call. The wrapper needs the prompt string to label the CB event trace, so a completion invoked without a prompt cannot be instrumented and fails here.

Source

Thrown at llama-index-core/llama_index/core/llms/callbacks.py:308

def llm_completion_callback() -> Callable:
    def wrap(f: Callable) -> Callable:
        @contextmanager
        def wrapper_logic(_self: Any) -> Generator[CallbackManager, None, None]:
            callback_manager = getattr(_self, "callback_manager", None)
            if not isinstance(callback_manager, CallbackManager):
                _self.callback_manager = CallbackManager()

            yield _self.callback_manager

        def extract_prompt(*args: Any, **kwargs: Any) -> str:
            if len(args) > 0:
                return str(args[0])
            elif "prompt" in kwargs:
                return kwargs["prompt"]
            else:
                raise ValueError(
                    "No prompt provided in positional or keyword arguments"
                )

        async def wrapped_async_llm_predict(
            _self: Any, *args: Any, **kwargs: Any
        ) -> Any:
            prompt = extract_prompt(*args, **kwargs)
            with (
                wrapper_logic(_self) as callback_manager,
                callback_manager.as_trace("completion"),
            ):
                span_id = active_span_id.get()
                model_dict = _self.to_payload()
                dispatcher.event(
                    LLMCompletionStartEvent(
                        model_dict=model_dict,
                        prompt=prompt,
                        additional_kwargs=kwargs,

View on GitHub (pinned to afd0fef371)

Solutions

1. Pass the prompt positionally: llm.complete('What is 2+2?') or as llm.complete(prompt='...').
2. For message lists, use llm.chat(messages=[...]) instead of complete().
3. Audit wrapper/dispatcher code that forwards kwargs to ensure 'prompt' is always present.

Example fix

# before
resp = llm.complete(messages=[ChatMessage('hi')])  # no 'prompt' arg -> ValueError

# after
resp = llm.chat(messages=[ChatMessage('hi')])
# or for plain completion:
resp = llm.complete(prompt='hi')

def call_complete(llm, *args, **kwargs):
    if not args and 'prompt' not in kwargs:
        raise ValueError('complete() requires a prompt')
    return llm.complete(*args, **kwargs)

When it happens

Trigger: Calling a decorated llm.complete() with no arguments, with only non-prompt kwargs (e.g. llm.complete(temperature=0)), or with the prompt under a different keyword (e.g. llm.complete(input='...', ) or messages=[...]) — chat-style calls routed through the completion wrapper.

Common situations: Generic dispatcher code that forwards **kwargs to complete() and sometimes passes nothing; refactoring chat() calls into complete() while keeping a messages kwarg; mocked LLMs in tests called with empty signatures that now hit the instrumented wrapper.

Related errors

• query and response must be provided
• Got empty streaming response
• Expected ActionReasoningStep, got {reasoning_step}
• This query engine does not support retrieve, use query direc
• This query engine does not support synthesize, use query dir

AI-assisted analysis of run-llama/llama_index@afd0fef371 (2026-08-15). Data as JSON: /api/errors/f97e26d2b70f559c. Report an issue: GitHub.