Builder for Milvus collection field schemas.
Provides a fluent API for constructing field definitions with validation. Supports all Milvus data types including scalars, vectors, and complex types.
# Primary key field (using builder)
field = Field.new("id", :int64) |> Field.set_primary_key() |> Field.auto_id()
# Vector field with dimension
field = Field.new("embedding", :float_vector) |> Field.dimension(128)
# VarChar with max length
field = Field.new("title", :varchar) |> Field.max_length(512)
# Using smart constructors
Field.primary_key("id", :int64, auto_id: true)
Field.vector("embedding", 128)
Field.varchar("title", 512, nullable: true)Creates an array field with the specified element type.
Checks if the field is an array of struct type.
Enables auto-generation of IDs for this field. Only valid for primary key fields with int64 or varchar type.
Marks the field as a clustering key.
Returns list of all supported data types.
Sets a default value for the field.
Sets the field description.
Sets the dimension for vector fields. Required for all vector types except sparse vectors.
Marks the field as a dynamic field.
Sets the element type for array fields.
Enables text analyzer for varchar fields. Used for BM25 full-text search.
Creates a Field from a protobuf FieldSchema struct.
Creates a Field from a StructArrayFieldSchema proto.
Sets the maximum capacity for array fields.
Sets the maximum length for varchar fields. Required for varchar type, must be between 1 and 65535.
Creates a new field with the given name and data type.
Marks the field as nullable.
Marks the field as a partition key.
Creates a primary key field with common defaults.
Creates a scalar field of the specified type.
Checks if the given type is a scalar type.
Returns list of scalar data types.
Marks the field as the primary key (sets to true).
Marks the field as the primary key.
Creates a sparse vector field.
Creates a struct field with nested fields.
Checks if the field is a struct type.
Creates a timestamp field (with timezone).
Converts the field to a protobuf FieldSchema struct.
Converts an array_of_struct field to a StructArrayFieldSchema proto.
Validates the field configuration.
Validates the field and raises on error.
Creates a varchar field with the specified max length.
Creates a vector field with the specified dimension.
Checks if the given type is a vector type.
Returns list of vector data types.
@type data_type() ::
:bool
| :int8
| :int16
| :int32
| :int64
| :float
| :double
| :varchar
| :json
| :text
| :timestamp
| :array
| :struct
| :array_of_struct
| :binary_vector
| :float_vector
| :float16_vector
| :bfloat16_vector
| :sparse_float_vector
| :int8_vector
@type t() :: %Milvex.Schema.Field{ auto_id: boolean(), data_type: data_type(), default_value: term() | nil, description: String.t() | nil, dimension: pos_integer() | nil, element_type: data_type() | nil, enable_analyzer: boolean(), is_clustering_key: boolean(), is_dynamic: boolean(), is_partition_key: boolean(), is_primary_key: boolean(), max_capacity: pos_integer() | nil, max_length: pos_integer() | nil, name: String.t(), nullable: boolean(), struct_schema: [t()] | nil }
Creates an array field with the specified element type.
:max_capacity - Maximum number of elements (required):nullable - Allow null values (default: false):description - Field description:struct_schema - Required when element_type is :struct, list of Field.t():max_length - For varchar element types# Array of scalars
Field.array("tags", :varchar, max_capacity: 100, max_length: 64)
# Array of structs
struct_fields = [
Field.varchar("text", 4096),
Field.vector("embedding", 1024)
]
Field.array("sentences", :struct, max_capacity: 50, struct_schema: struct_fields)Checks if the field is an array of struct type.
Enables auto-generation of IDs for this field. Only valid for primary key fields with int64 or varchar type.
Marks the field as a clustering key.
@spec data_types() :: [data_type()]
Returns list of all supported data types.
Sets a default value for the field.
Sets the field description.
@spec dimension(t(), pos_integer()) :: t()
Sets the dimension for vector fields. Required for all vector types except sparse vectors.
Marks the field as a dynamic field.
Only valid for scalar types: bool, int8, int16, int32, int64, float, double, text, varchar, json. Dynamic fields allow storing data that may not be strictly typed at the schema level.
Field.new("metadata", :json) |> Field.dynamic()
Field.varchar("extra", 256, dynamic: true)Sets the element type for array fields.
Enables text analyzer for varchar fields. Used for BM25 full-text search.
@spec from_proto(Milvex.Milvus.Proto.Schema.FieldSchema.t()) :: t()
Creates a Field from a protobuf FieldSchema struct.
@spec from_struct_array_field_schema( Milvex.Milvus.Proto.Schema.StructArrayFieldSchema.t() ) :: t()
Creates a Field from a StructArrayFieldSchema proto.
@spec max_capacity(t(), pos_integer()) :: t()
Sets the maximum capacity for array fields.
@spec max_length(t(), pos_integer()) :: t()
Sets the maximum length for varchar fields. Required for varchar type, must be between 1 and 65535.
Creates a new field with the given name and data type.
name - Field name (1-255 characters, alphanumeric and underscores)data_type - One of the supported Milvus data typesField.new("id", :int64)
Field.new("embedding", :float_vector)Marks the field as nullable.
Marks the field as a partition key.
Creates a primary key field with common defaults.
:auto_id - Enable auto ID generation (default: false):description - Field descriptionField.primary_key("id", :int64)
Field.primary_key("id", :int64, auto_id: true)
Field.primary_key("pk", :varchar, max_length: 64)Creates a scalar field of the specified type.
:nullable - Allow null values (default: false):description - Field description:default - Default valueField.scalar("age", :int32)
Field.scalar("score", :float, nullable: true)Checks if the given type is a scalar type.
@spec scalar_types() :: [data_type()]
Returns list of scalar data types.
Marks the field as the primary key (sets to true).
Marks the field as the primary key.
When called with a single Field argument, sets is_primary_key to true.
When called with a boolean, sets is_primary_key to that value.
Creates a sparse vector field.
Sparse vectors don't require a dimension parameter.
Field.sparse_vector("sparse_embedding")Creates a struct field with nested fields.
Struct fields are typically used as the element type for array_of_struct fields.
:fields - List of Field.t() defining the struct schema (required):description - Field descriptionstruct_fields = [
Field.varchar("text", 4096),
Field.vector("embedding", 1024),
Field.varchar("speaker_id", 36)
]
Field.struct("sentence", fields: struct_fields)Checks if the field is a struct type.
Creates a timestamp field (with timezone).
Milvus stores timestamps as Timestamptz - timestamp with timezone information.
:nullable - Allow null values (default: false):description - Field description:default - Default valueField.timestamp("created_at")
Field.timestamp("updated_at", nullable: true)@spec to_proto(t()) :: Milvex.Milvus.Proto.Schema.FieldSchema.t()
Converts the field to a protobuf FieldSchema struct.
@spec to_struct_array_field_schema(t()) :: Milvex.Milvus.Proto.Schema.StructArrayFieldSchema.t()
Converts an array_of_struct field to a StructArrayFieldSchema proto.
Used for the struct_array_fields in CollectionSchema.
Nested fields are wrapped as Array (for scalars) or ArrayOfVector (for vectors) as required by Milvus's StructArrayFieldSchema format.
@spec validate(t()) :: {:ok, t()} | {:error, Milvex.Error.t()}
Validates the field configuration.
Returns {:ok, field} if valid, {:error, error} otherwise.
Validates the field and raises on error.
@spec varchar(String.t(), pos_integer(), keyword()) :: t()
Creates a varchar field with the specified max length.
:nullable - Allow null values (default: false):description - Field description:default - Default value:enable_analyzer - Enable text analyzer for BM25 search (default: false)Field.varchar("title", 256)
Field.varchar("description", 1024, nullable: true)
Field.varchar("content", 4096, enable_analyzer: true)@spec vector(String.t(), pos_integer(), keyword()) :: t()
Creates a vector field with the specified dimension.
:type - Vector type (default: :float_vector):description - Field descriptionField.vector("embedding", 128)
Field.vector("embedding", 768, type: :float16_vector)Checks if the given type is a vector type.
@spec vector_types() :: [data_type()]
Returns list of vector data types.