microsoft/autogen · error · ValueError

Please provide model_path if ... or provide repo_id and file

Error message

Please provide model_path if ... or provide repo_id and filename if ....

What it means

LlamaCppChatCompletionClient's constructor requires exactly one way to locate a model: either a model_path kwarg passed straight to Llama(...), or both repo_id and filename (both non-empty) passed to Llama.from_pretrained(...). If kwargs contain none of these, the final else branch raises this ValueError. It is a configuration-validation error fired at client construction time.

Source

Thrown at python/packages/autogen-ext/src/autogen_ext/models/llama_cpp/_llama_cpp_completion_client.py:260

        if model_info:
            validate_model_info(model_info)
            self._model_info = model_info
        else:
            # Default model info.
            self._model_info = self.DEFAULT_MODEL_INFO

        if "repo_id" in kwargs and "filename" in kwargs and kwargs["repo_id"] and kwargs["filename"]:
            repo_id: str = cast(str, kwargs.pop("repo_id"))
            filename: str = cast(str, kwargs.pop("filename"))
            pretrained = Llama.from_pretrained(repo_id=repo_id, filename=filename, **kwargs)  # type: ignore
            assert isinstance(pretrained, Llama)
            self.llm = pretrained

        elif "model_path" in kwargs:
            self.llm = Llama(**kwargs)  # pyright: ignore[reportUnknownMemberType]
        else:
            raise ValueError("Please provide model_path if ... or provide repo_id and filename if ....")
        self._total_usage = {"prompt_tokens": 0, "completion_tokens": 0}

    async def create(
        self,
        messages: Sequence[LLMMessage],
        *,
        tools: Sequence[Tool | ToolSchema] = [],
        tool_choice: Tool | Literal["auto", "required", "none"] = "auto",
        # None means do not override the default
        # A value means to override the client default - often specified in the constructor
        json_output: Optional[bool | type[BaseModel]] = None,
        extra_create_args: Mapping[str, Any] = {},
        cancellation_token: Optional[CancellationToken] = None,
    ) -> CreateResult:
        create_args = dict(extra_create_args)
        # Convert LLMMessage objects to dictionaries with 'role' and 'content'
        # converted_messages: List[Dict[str, str | Image | list[str | Image] | list[FunctionCall]]] = []
        converted_messages: list[

View on GitHub (pinned to 027ecf0a37)

Solutions

1. Pass a local GGUF file path: LlamaCppChatCompletionClient(model_path='/path/to/model.gguf')
2. Or pull from Hugging Face with BOTH arguments: LlamaCppChatCompletionClient(repo_id='Qwen/Qwen2-0.5B-Instruct-GGUF', filename='qwen2-0_5b-instruct-q4_k_m.gguf')
3. Check spelling and case of the kwarg — it must be exactly model_path (or the pair repo_id/filename); any extra misspelled keys are silently forwarded to Llama() as unused kwargs
4. If building kwargs dynamically, assert 'model_path' in kwargs or ('repo_id' in kwargs and 'filename' in kwargs) before constructing the client

Example fix

# before
client = LlamaCppChatCompletionClient(model='my-model.gguf')  # wrong kwarg name

# after
client = LlamaCppChatCompletionClient(model_path='my-model.gguf')

def assert_llama_config(kwargs: dict) -> None:
    has_path = "model_path" in kwargs
    has_repo = bool(kwargs.get("repo_id")) and bool(kwargs.get("filename"))
    if not (has_path or has_repo):
        raise ValueError("Supply model_path, or both repo_id and filename")

assert_llama_config(client_kwargs)
client = LlamaCppChatCompletionClient(**client_kwargs)

def has_valid_model_source(kwargs: Mapping[str, Any]) -> bool:
    return (
        "model_path" in kwargs
        or (bool(kwargs.get("repo_id")) and bool(kwargs.get("filename")))
    )

try:
    client = LlamaCppChatCompletionClient(**cfg)
except ValueError as e:
    if "model_path" in str(e):
        raise SystemExit(f"Bad model config: {e}") from e
    raise

When it happens

Trigger: Calling LlamaCppChatCompletionClient() with no model_path, repo_id, or filename; passing repo_id alone or filename alone (the branch requires both keys present AND truthy); passing model_path=None or empty string only if the key is absent — note only key presence is checked for model_path ('model_path' in kwargs).

Common situations: Copied an example for a different model client and forgot the local GGUF path; typo like modelpath='...' or path_to_model=... so the key never reaches kwargs; assumed repo_id alone is enough to pull from Hugging Face; env var for the model path empty so the key was never set.

Related errors

• Unsupported config type {config.GetType()}
• Cannot save screenshots without a debug directory. Set it us
• Timeout must be greater than or equal to 1.
• endpoint is required for AzureAIChatCompletionClient
• credential is required for AzureAIChatCompletionClient

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/cb821d2b5039ea90. Report an issue: GitHub.