chroma-core/chroma · error · ValueError

Preferred providers must be unique

Error message

Preferred providers must be unique

What it means

The second constructor check in ONNXMiniLM_L6_V2 compares len(preferred_providers) with len(set(preferred_providers)); duplicates mean the same execution provider would be registered twice in the ONNX InferenceSession, which is meaningless and usually signals a config-generation bug, so it raises ValueError("Preferred providers must be unique"). Order still matters (first match wins in ORT), so dedupe must preserve order.

Source

Thrown at chromadb/utils/embedding_functions/onnx_mini_lm_l6_v2.py:64

    def __init__(self, preferred_providers: Optional[List[str]] = None) -> None:
        """
        Initialize the ONNXMiniLM_L6_V2 embedding function.

        Args:
            preferred_providers (List[str], optional): The preferred ONNX runtime providers.
                Defaults to None.
        """
        # convert the list to set for unique values
        if preferred_providers and not all(
            [isinstance(i, str) for i in preferred_providers]
        ):
            raise ValueError("Preferred providers must be a list of strings")
        # check for duplicate providers
        if preferred_providers and len(preferred_providers) != len(
            set(preferred_providers)
        ):
            raise ValueError("Preferred providers must be unique")

        self._preferred_providers = preferred_providers

        try:
            # Equivalent to import onnxruntime
            self.ort = importlib.import_module("onnxruntime")
        except ImportError:
            raise ValueError(
                "The onnxruntime python package is not installed. Please install it with `pip install onnxruntime`"
            )
        try:
            # Equivalent to from tokenizers import Tokenizer
            self.Tokenizer = importlib.import_module("tokenizers").Tokenizer
        except ImportError:
            raise ValueError(
                "The tokenizers python package is not installed. Please install it with `pip install tokenizers`"
            )
        try:

View on GitHub (pinned to aecdd12c8a)

Solutions

1. Dedupe while preserving order before constructing: seen=set(); providers=[p for p in providers if not (p in seen or seen.add(p))]
2. Fix the config merge that produces the duplicate (don't append the default CPU fallback if it is already present)
3. Drop the argument entirely (pass None) to let the EF use all available providers

Example fix

// before
fn = ONNXMiniLM_L6_V2(preferred_providers=["CUDAExecutionProvider", "CPUExecutionProvider", "CPUExecutionProvider"])  # ValueError

// after
providers = list(dict.fromkeys(["CUDAExecutionProvider", "CPUExecutionProvider", "CPUExecutionProvider"]))
fn = ONNXMiniLM_L6_V2(preferred_providers=providers)

providers = list(dict.fromkeys(providers)) if providers else providers  # order-preserving dedupe
fn = ONNXMiniLM_L6_V2(preferred_providers=providers)

When it happens

Trigger: ONNXMiniLM_L6_V2(preferred_providers=["CPUExecutionProvider", "CPUExecutionProvider"]); concatenating a user provider list with a default fallback list without deduping (e.g. cfg_providers + ["CPUExecutionProvider"]); config templating that injects the same provider for GPU and CPU sections.

Common situations: Appending CPUExecutionProvider as a fallback after user-specified providers that already include it; YAML anchors reusing a provider list twice; environment-specific overrides merged on top of defaults with the same entry.

Related errors

• Preferred providers must be a list of strings
• Preferred providers must be subset of available providers: {
• Could not build embedding function {ef_config['name']} from
• Updating '{key}' is not supported for {NAME}
• The onnxruntime python package is not installed. Please inst

AI-assisted analysis of chroma-core/chroma@aecdd12c8a (2026-08-16). Data as JSON: /api/errors/3256084b53428a6a. Report an issue: GitHub.