microsoft/graphrag · error · ValueError

Each item in ModelConfig.mock_responses must be a float.

Error message

Each item in ModelConfig.mock_responses must be a float.

What it means

MockLLMEmbedding validates every entry of ModelConfig.mock_responses is a float (integers are rejected too, due to the strict isinstance check). This guarantees the mock returns well-formed embedding vectors.

Source

Thrown at packages/graphrag-llm/graphrag_llm/embedding/mock_llm_embedding.py:52

        self,
        *,
        model_config: "ModelConfig",
        tokenizer: "Tokenizer",
        metrics_store: "MetricsStore",
        **kwargs: Any,
    ):
        """Initialize MockLLMEmbedding."""
        self._tokenizer = tokenizer
        self._metrics_store = metrics_store

        mock_responses = model_config.mock_responses
        if not isinstance(mock_responses, list) or len(mock_responses) == 0:
            msg = "ModelConfig.mock_responses must be a non-empty list of embedding responses."
            raise ValueError(msg)

        if not all(isinstance(resp, float) for resp in mock_responses):
            msg = "Each item in ModelConfig.mock_responses must be a float."
            raise ValueError(msg)

        self._mock_responses = mock_responses  # type: ignore

    def embedding(
        self, /, **kwargs: Unpack["LLMEmbeddingArgs"]
    ) -> "LLMEmbeddingResponse":
        """Sync embedding method."""
        input = kwargs.get("input")
        response = create_embedding_response(
            self._mock_responses, batch_size=len(input)
        )
        self._mock_index += 1
        return response

    async def embedding_async(
        self, /, **kwargs: Unpack["LLMEmbeddingArgs"]
    ) -> "LLMEmbeddingResponse":
        """Async embedding method."""

View on GitHub (pinned to f40e9a26ce)

Solutions

1. Ensure every element is a float literal, e.g. [0.1, 0.0, 1.0] (use 1.0 not 1 in YAML/JSON)
2. If values come from external data, coerce with [float(x) for x in values] before building ModelConfig

Example fix

# before
ModelConfig(type=LLMProviderType.MockLLM, mock_responses=[1, 0, 1])
# after
ModelConfig(type=LLMProviderType.MockLLM, mock_responses=[1.0, 0.0, 1.0])

cfg_dict["mock_responses"] = [float(x) for x in raw_responses]  # coerce before ModelConfig(**cfg_dict)

def is_float_list(v) -> bool:
    return isinstance(v, list) and len(v) > 0 and all(isinstance(x, float) for x in v)

When it happens

Trigger: mock_responses containing strings (["0.1"]), ints ([1, 2]), None, or nested lists when constructing MockLLMEmbedding.

Common situations: YAML mock_responses: [1, 0, 1] parsed as ints; JSON config with string numbers; copy-pasted fixture data of the wrong type.

Related errors

• ModelConfig.mock_responses must be a non-empty list of embed
• ModelConfig.type '{strategy}' is not registered in the Compl
• request_id needs to be passed as a keyword argument
• Reports missing {source_col} column
• Response must be a list of dictionaries.

AI-assisted analysis of microsoft/graphrag@f40e9a26ce (2026-08-27). Data as JSON: /api/errors/396f855b332d2547. Report an issue: GitHub.