lancedb/lancedb · error · TypeError

Unknown data type . Supported types: list of dicts, pandas…

Error message

Unknown data type {type(data)}. Supported types: list of dicts, pandas DataFrame, polars DataFrame, pyarrow Table/RecordBatch, or Pydantic models. See https://docs.lancedb.com/tables/ for examples.

What it means

data_to_reader dispatches on the type of `data` and raises TypeError when the value matches none of the supported input types. LanceDB only converts list-of-dicts, pandas DataFrames, polars DataFrames, pyarrow Table/RecordBatch/RecordBatchReader, Pydantic models, and generic Iterables (with a schema). Anything else (str, int, bytes, ndarray, dict, etc.) is rejected.

Solutions

1. Convert to a supported type first: pa.Table.from_pandas(df), pd.DataFrame(rows), or a list of dicts
2. For a NumPy array, wrap it: pa.table({'col': arr}) or pd.DataFrame(arr)
3. For a single dict, wrap it in a list: db.create_table('t', data=[row])
4. For other frameworks (HuggingFace Datasets, Dask), convert to arrow/pandas before passing
5. If it's an Iterable, supply schema=<pyarrow schema> so data_to_reader takes the iterable branch

Example fix

// before
import numpy as np
arr = np.random.rand(10, 8)
db.create_table('t', data=arr)  # TypeError

// after
import pyarrow as pa
db.create_table('t', data=pa.table({'vec': arr.tolist()}))

SUPPORTED = (list, 'pandas.DataFrame', 'polars.DataFrame', 'pyarrow.Table', 'pyarrow.RecordBatch', 'pyarrow.RecordBatchReader')
import pyarrow as pa
ok = isinstance(data, (list, pa.Table, pa.RecordBatch, pa.RecordBatchReader)) or type(data).__name__ in ('DataFrame',)

def is_supported_data(data) -> bool:
    import pyarrow as pa
    if isinstance(data, (list, pa.Table, pa.RecordBatch, pa.RecordBatchReader)):
        return True
    mod = type(data).__module__
    return type(data).__name__ == 'DataFrame' and mod.startswith(('pandas', 'polars'))

try:
    db.create_table('t', data=data)
except TypeError as e:
    if 'Unknown data type' in str(e):
        db.create_table('t', data=pa.Table.from_pandas(pd.DataFrame(data)))

When it happens

Trigger: Passing an unsupported object to create_table's data parameter, e.g. `db.create_table('t', data=multiline_string)`, a single dict instead of a list, a NumPy array, or a dict-of-lists.

Common situations: Copy-pasting CSV/JSON text as data; passing a dict of column arrays from another framework; accidentally passing a file path string instead of reading it; passing a HuggingFace Dataset or Dask/Spark frame not supported by this overload.

Understand the failure class

Background: Type mismatch errors: IllegalArgumentException, TypeError and type guards across 150 open-source libraries — this error's family across 150 libraries.

Related errors

• columns must be a list of column names or a dict
• config must be an instance of IvfSq, IvfPq, IvfRq, HnswPq…
• create_function_async requires a @udf definition
• dict values must be str or Expr, got
• Each input should be either str, bytes, Path or Image.

AI-assisted analysis of lancedb/lancedb@c7b051aff7 (2026-09-08). Data as JSON: /api/errors/da7b09cf5aae6c4e. Report an issue: GitHub.

Appendix: source

Thrown at python/python/lancedb/common.py:95

        return data.to_reader()
    elif isinstance(data, pa.RecordBatchReader):
        return data
    elif (
        type(data).__module__.startswith("polars")
        and data.__class__.__name__ == "DataFrame"
    ):
        return data.to_arrow().to_reader()
    # for other iterables, assume they are of type Iterable[RecordBatch]
    elif isinstance(data, Iterable):
        if schema is not None:
            data = _casting_recordbatch_iter(data, schema)
            return pa.RecordBatchReader.from_batches(schema, data)
        else:
            raise ValueError(
                "Must provide schema to write dataset from RecordBatch iterable"
            )
    else:
        raise TypeError(
            f"Unknown data type {type(data)}. "
            "Supported types: list of dicts, pandas DataFrame, polars DataFrame, "
            "pyarrow Table/RecordBatch, or Pydantic models. "
            "See https://docs.lancedb.com/tables/ for examples."
        )


def validate_schema(schema: pa.Schema):
    """
    Make sure the metadata is valid utf8
    """
    if schema.metadata is not None:
        _validate_metadata(schema.metadata)


def _validate_metadata(metadata: dict):
    """
    Make sure the metadata values are valid utf8 (can be nested)

View on GitHub (pinned to c7b051aff7)