crewAIInc/crewAI · error · ValueError

Invalid data_type: '{raw_data_type}'. Valid values are: 'fil

Error message

Invalid data_type: '{raw_data_type}'. Valid values are: 'file' (auto-detect), or one of: {', '.join(dt.value for dt in DataType)}

What it means

ValueError raised while parsing the data_type argument passed to the RAG adapter's add-content API. String values are mapped through the DataType enum; the sentinel string 'file' means auto-detect and is skipped, and passing a DataType instance is accepted. Any other string that is not an enum value raises this error listing the valid options.

Source

Thrown at lib/crewai-tools/src/crewai_tools/adapters/crewai_rag_adapter.py:159

        from crewai_tools.rag.base_loader import LoaderResult
        from crewai_tools.rag.data_types import DataType, DataTypes
        from crewai_tools.rag.source_content import SourceContent

        documents: list[BaseRecord] = []
        raw_data_type = kwargs.get("data_type")
        base_metadata: dict[str, Any] = kwargs.get("metadata", {})

        data_type: DataType | None = None
        if raw_data_type is not None:
            if isinstance(raw_data_type, DataType):
                if raw_data_type != DataType.FILE:
                    data_type = raw_data_type
            elif isinstance(raw_data_type, str):
                if raw_data_type != "file":
                    try:
                        data_type = DataType(raw_data_type)
                    except ValueError:
                        raise ValueError(
                            f"Invalid data_type: '{raw_data_type}'. "
                            f"Valid values are: 'file' (auto-detect), or one of: "
                            f"{', '.join(dt.value for dt in DataType)}"
                        ) from None

        content_items: list[ContentItem] = list(args)

        path_value = kwargs.get("path") or kwargs.get("file_path")
        if path_value is not None:
            content_items.append(path_value)

        if url := kwargs.get("url"):
            content_items.append(url)
        if website := kwargs.get("website"):
            content_items.append(website)
        if github_url := kwargs.get("github_url"):
            content_items.append(github_url)
        if youtube_url := kwargs.get("youtube_url"):

View on GitHub (pinned to 754d7323be)

Solutions

1. Use the enum directly: from the data types module, pass DataType.PDF_FILE (or its exact string value) instead of guessing.
2. Omit data_type entirely to use auto-detection from the content/extension.
3. Print valid values once to discover exact names: [dt.value for dt in DataType].

Example fix

# before
adapter("report.pdf", data_type="pdf")  # ValueError: Invalid data_type

# after
adapter("report.pdf", data_type=DataType.PDF_FILE)
# or simply omit for auto-detection:
adapter("report.pdf")

from crewai_tools.adapters.crewai_rag_adapter import DataType  # adjust to real export

valid = {dt.value for dt in DataType}
if isinstance(data_type, str) and data_type != "file" and data_type not in valid:
    raise ValueError(f"bad data_type {data_type!r}; choose from {sorted(valid)}")

def is_valid_data_type(value: object) -> bool:
    if isinstance(value, DataType):
        return True
    return isinstance(value, str) and (value == "file" or value in {dt.value for dt in DataType})

When it happens

Trigger: Calling the adapter with data_type='pdf_file' when the enum value is actually 'pdf_file' vs 'PDF_FILE' mismatch, or a typo like 'pdf', 'doc', 'csv_file', 'website' that is not a DataType member.

Common situations: Guessing type names instead of checking DataType values; version drift when enum member names changed between releases; passing 'file' variants like 'files' or 'auto'.

Related errors

• Directory does not exist: {source_ref}
• Client is not initialized
• File does not exist: {source_ref}
• Chunk overlap ({chunk_overlap}) cannot be >= chunk size ({ch
• No chunker defined for {self}

AI-assisted analysis of crewAIInc/crewAI@754d7323be (2026-08-15). Data as JSON: /api/errors/d22aebd7f5e4b80f. Report an issue: GitHub.