FieldSchema

Defines the name, data type, and field-specific options used in a collection schema.

Request Syntax

FieldSchema(
    name: str,
    dtype: DataType,
    description: str = "",
    **kwargs
)

PARAMETERS:

• name (str) - [REQUIRED] Name of the field.

• dtype (DataType) - [REQUIRED] Data type of the field.

• description (str) - Default: "" Description of the field.

• kwargs (Any) - Additional field options.

  • is_primary (bool) - Whether the field is the collection primary key.

  • auto_id (bool) - Whether Milvus automatically generates primary-key values.

  • nullable (bool) - Whether the field can contain null values.

  • default_value (Any) - Default value used when the field value is omitted.

  • dim (int) - Dimension of a dense vector field.

  • max_length (int) - Maximum length of a VARCHAR field.

  • element_type (DataType) - Data type of elements in an ARRAY field.

  • is_partition_key (bool) - Whether the field is the partition key.

  • is_clustering_key (bool) - Whether the field is the clustering key.

RETURN TYPE:

FieldSchema

RETURNS:

Field schema instance containing the configured data type, constraints, default, and nullable metadata.

EXCEPTIONS:

• DataTypeNotSupportException Raised when dtype is not a supported DataType value. Use a supported PyMilvus data type.

• PrimaryKeyException Raised when primary-key or auto-ID options are invalid. Check the is_primary and auto_id option types and combinations.

• PartitionKeyException Raised when is_partition_key is not a Boolean value. Pass True or False for the partition-key option.

• ClusteringKeyException Raised when is_clustering_key is not a Boolean value. Pass True or False for the clustering-key option.

• ParamError Raised when a field option or default value is invalid. Check the options required by the selected data type.

Examples

from pymilvus import CollectionSchema, DataType, FieldSchema

schema = CollectionSchema(
    fields=[
        FieldSchema(name="id", dtype=DataType.INT64, is_primary=True),
        FieldSchema(name="vector", dtype=DataType.FLOAT_VECTOR, dim=3),
    ]
)
print(schema)