package milvus.proto.schema

Get desktop application:
View/edit binary Protocol Buffers messages

Used in: SearchResultData

• repeated BucketKeyEntry key = 1

• int64 count = 2

• map<string, MetricValue> metrics = 3

  metrics carries a typed metric result per alias. min/max may yield string/bool values when the source field has those types; count/sum/avg stay numeric. The MetricValue oneof avoids a lossy all-double encoding.

• repeated AggHit hits = 4

• repeated AggBucket sub_groups = 5

Used in: AggBucket

• oneof pk

  • int64 int_pk = 1

  • string str_pk = 2

• float score = 3

• repeated AggHitField fields = 4

Used in: AggHit

• int64 field_id = 1

• string field_name = 2

• oneof value

  • int64 int_val = 3

  • bool bool_val = 4

  • float float_val = 5

  • double double_val = 6

  • string string_val = 7

  • bytes bytes_val = 8

Used in: ScalarField

• repeated ScalarField data = 1

• DataType element_type = 2

Used in: ScalarField, TemplateArrayValue

• repeated bool data = 1

Used in: AggBucket

• int64 field_id = 1

• string field_name = 2

• oneof value

  • int64 int_val = 3

  • string string_val = 4

  • bool bool_val = 5

For special fields such as bigdecimal, array...

Used in: ScalarField

• repeated bytes data = 1

clustering distribution info of a certain data unit, it can be segment, partition, etc.

• repeated VectorClusteringInfo vector_clustering_infos = 1

• repeated ScalarClusteringInfo scalar_clustering_infos = 2

* @brief Collection schema

Used in: milvus.DescribeCollectionResponse, msg.CreateCollectionRequest, msg.ImportMsg

• string name = 1

• string description = 2

• bool autoID = 3

  deprecated later, keep compatible with c++ part now

• repeated FieldSchema fields = 4

• bool enable_dynamic_field = 5

  mark whether this table has the dynamic field function enabled.

• repeated common.KeyValuePair properties = 6

• repeated FunctionSchema functions = 7

• string dbName = 8

• repeated StructArrayFieldSchema struct_array_fields = 9

• int32 version = 10

  The version of current collection schema updates, increment every time the schema content is updating,

• string external_source = 11

  0 when the collection is created or the collection schema is updated before 2.6.5. the name dbName and description take no effect to it.

• string external_spec = 12

• bool do_physical_backfill = 13

• repeated int64 file_resource_ids = 14

• bool enable_namespace = 15

* @brief Field data type

Used in: ArrayArray, FieldData, FieldSchema, VectorArray

• None = 0

• Bool = 1

• Int8 = 2

• Int16 = 3

• Int32 = 4

• Int64 = 5

• Float = 10

• Double = 11

• String = 20

• VarChar = 21

  variable-length strings with a specified maximum length

• Array = 22

• JSON = 23

• Geometry = 24

• Text = 25

• Timestamptz = 26

• Mol = 27

  Molecular data type

• BinaryVector = 100

• FloatVector = 101

• Float16Vector = 102

• BFloat16Vector = 103

• SparseFloatVector = 104

• Int8Vector = 105

• ArrayOfVector = 106

• ArrayOfStruct = 200

  internal types, not used in user interface

• Struct = 201

Used in: ScalarField, TemplateArrayValue

• repeated double data = 1

Used in: milvus.InsertRequest, milvus.QueryResults, milvus.UpsertRequest, msg.InsertRequest, SearchResultData, StructArrayField

• DataType type = 1

• string field_name = 2

• oneof field

  • ScalarField scalars = 3

  • VectorField vectors = 4

  • StructArrayField struct_arrays = 8

• int64 field_id = 5

• bool is_dynamic = 6

• repeated bool valid_data = 7

FieldPartialUpdateOp describes how the values carried by the matching FieldData should be applied against the existing row during a partial upsert. The message is referenced from UpsertRequest.field_ops rather than embedded in FieldData to keep FieldData a pure data carrier — FieldData flows through InsertRequest, QueryResults, SearchResultData and internal msgstream paths where an op directive would be meaningless and risk accidental echo-back on client-side read-modify-write flows. Ops are matched to FieldData entries by field_name. When no op message targets a given field, that field is merged with REPLACE semantics (full overwrite), preserving backward compatibility.

Used in: milvus.UpsertRequest

• string field_name = 1

  field_name identifies the FieldData this op targets. Must match a field carried by UpsertRequest.fields_data; unknown field names are rejected by the server.

• FieldPartialUpdateOp.OpType op = 2

Used in: FieldPartialUpdateOp

• REPLACE = 0

  REPLACE overwrites the field with the new values. Default behavior when no op targets a field. Applicable to all field types.

• ARRAY_APPEND = 1

  ARRAY_APPEND appends the new values to the tail of the existing array. Requires DataType_Array. The resulting length must not exceed the field's max_capacity.

• ARRAY_REMOVE = 2

  ARRAY_REMOVE removes every occurrence of each provided value from the existing array. Requires DataType_Array. No-op when the base array is empty or no element matches.

* @brief Field schema

Used in: milvus.AlterCollectionSchemaRequest.FieldInfo, CollectionSchema, StructArrayFieldSchema

• int64 fieldID = 1

• string name = 2

• bool is_primary_key = 3

• string description = 4

• DataType data_type = 5

• repeated common.KeyValuePair type_params = 6

• repeated common.KeyValuePair index_params = 7

• bool autoID = 8

• FieldState state = 9

  To keep compatible with older version, the default

• DataType element_type = 10

  state is `Created`.

  For array type, the element type is stored here

• optional ValueField default_value = 11

  default_value only support scalars except array, json and timestamptz for now

• bool is_dynamic = 12

  mark whether this field is the dynamic field

• bool is_partition_key = 13

  enable logic partitions

• bool is_clustering_key = 14

• bool nullable = 15

  enable set null as field value

• bool is_function_output = 16

• string external_field = 17

  external field name

Used in: FieldSchema

• FieldCreated = 0

• FieldCreating = 1

• FieldDropping = 2

• FieldDropped = 3

Used in: milvus.CalcDistanceResults, ScalarField, VectorField

• repeated float data = 1

FunctionChain is an ordered public plan for function-based processing.

Used in: milvus.HybridSearchRequest, milvus.SearchRequest

• string name = 1

  Optional SDK/debug name.

• FunctionChainStage stage = 2

  Function chain execution stage.

• repeated FunctionChainOp ops = 3

  Ordered operations. The server executes ops in this order.

Used in: FunctionChainExprArg

• string name = 1

  Examples: "doc", "ts", "score1", "$score", "$seg_offset".

Used in: FunctionChainOp

• string name = 1

  Expression implementation name, e.g. "decay", "sum", "expr", "rerank_model", "udf_xxx". For fn.expr(col("$score") * 100), name is "expr" and the inner expression can be carried by params, for example params["expr"] = "$0 * 100".

• repeated FunctionChainExprArg args = 2

  Ordered expression arguments. Args may be collection fields, temporary values, system names, or literals. Function expr implementations interpret args by position.

• map<string, FunctionParamValue> params = 3

  Optional expression-level typed parameters.

Used in: FunctionChainExpr

• oneof arg

  • FunctionChainColumnArg column = 1

    Collection field, temporary variable, or system name.

  • FunctionParamValue literal = 2

    Literal argument.

Used in: FunctionChain

• string op = 1

  Operation name, e.g. "map", "filter", "sort", "limit", "merge".

• optional FunctionChainExpr expr = 2

  Optional function expression configured for this operation. Expr-based ops such as map/filter use this field. Non-expr ops such as sort/limit may leave it empty.

• repeated string inputs = 3

  Optional op-level input names. Mainly used by non-expr ops that still need caller-declared inputs, for example sort.inputs = ["$score"]. For expr-based ops, inputs are collected from expr.args column references.

• repeated string outputs = 4

  Optional operation output names. Outputs may be temporary variables or system names such as "$score". Writing collection fields is not supported.

• map<string, FunctionParamValue> params = 5

  Optional operation-level typed parameters.

beta, api may change Function chain APIs, including FunctionChainStage and related FunctionChain messages, are experimental and may change in future releases.

Used in: FunctionChain

• FunctionChainStageUnspecified = 0

• FunctionChainStageIngestion = 1

• FunctionChainStagePreProcess = 2

• FunctionChainStageL0Rerank = 3

• FunctionChainStageL1Rerank = 4

• FunctionChainStageL2Rerank = 5

• FunctionChainStagePostProcess = 6

Used in: FunctionParamValue

• repeated FunctionParamValue values = 1

Used in: FunctionParamValue

• map<string, FunctionParamValue> fields = 1

Used in: FunctionChainExpr, FunctionChainExprArg, FunctionChainOp, FunctionParamArray, FunctionParamObject

• oneof value

  • bool bool_value = 1

  • int64 int64_value = 2

  • double double_value = 3

  • string string_value = 4

  • FunctionParamArray array_value = 5

  • FunctionParamObject object_value = 6

  • bytes bytes_value = 7

Used in: milvus.AddCollectionFunctionRequest, milvus.AlterCollectionFunctionRequest, milvus.AlterCollectionSchemaRequest.AddRequest, CollectionSchema, FunctionScore

• string name = 1

• int64 id = 2

• string description = 3

• FunctionType type = 4

• repeated string input_field_names = 5

• repeated int64 input_field_ids = 6

• repeated string output_field_names = 7

• repeated int64 output_field_ids = 8

• repeated common.KeyValuePair params = 9

Used in: milvus.HybridSearchRequest, milvus.SearchRequest

• repeated FunctionSchema functions = 1

  Currently only a single function is supported, and multiple functions

• repeated common.KeyValuePair params = 2

  may be supported in the future, so repeated is used here.

Used in: FunctionSchema

• Unknown = 0

• BM25 = 1

• TextEmbedding = 2

• Rerank = 3

• MinHash = 4

• MolFingerprint = 5

Used in: ScalarField

• repeated bytes data = 1

Used in: ScalarField

• repeated string data = 1

Used in: milvus.MutationResult, milvus.SearchRequest, milvus.VectorIDs, msg.DeleteRequest, SearchResultData

• oneof id_field

  • LongArray int_id = 1

  • StringArray str_id = 2

Used in: milvus.CalcDistanceResults, ScalarField

• repeated int32 data = 1

Used in: ScalarField, TemplateArrayValue

• repeated bytes data = 1

Used in: milvus.ElementIndices, milvus.FlushCollectionResult, milvus.FlushResponse, IDs, ScalarField, SearchResultData, TemplateArrayValue

• repeated int64 data = 1

MetricValue is the typed result of a single metric aggregation.

Used in: AggBucket

• oneof value

  • int64 int_val = 1

  • double double_val = 2

  • string string_val = 3

  • bool bool_val = 4

MolArray stores processed molecular data in a serialized binary format.

Used in: ScalarField

• repeated bytes data = 1

MolSmilesArray stores user-provided SMILES strings for molecular data.

Used in: ScalarField

• repeated string data = 1

Scalar field clustering info todo more definitions: min/max, etc

Used in: ClusteringInfo

• string field = 1

Used in: ArrayArray, FieldData

• oneof data

  • BoolArray bool_data = 1

  • IntArray int_data = 2

  • LongArray long_data = 3

  • FloatArray float_data = 4

  • DoubleArray double_data = 5

  • StringArray string_data = 6

  • BytesArray bytes_data = 7

  • ArrayArray array_data = 8

  • JSONArray json_data = 9

  • GeometryArray geometry_data = 10

  • TimestamptzArray timestamptz_data = 11

  • GeometryWktArray geometry_wkt_data = 12

  • MolArray mol_data = 13

  • MolSmilesArray mol_smiles_data = 14

Used in: SearchResultData

• string token = 1

• float last_bound = 2

Used in: milvus.SearchResults

• int64 num_queries = 1

• int64 top_k = 2

• repeated FieldData fields_data = 3

• repeated float scores = 4

• optional IDs ids = 5

• repeated int64 topks = 6

• repeated string output_fields = 7

• optional FieldData group_by_field_value = 8

• int64 all_search_count = 9

• repeated float distances = 10

• optional SearchIteratorV2Results search_iterator_v2_results = 11

• repeated float recalls = 12

• string primary_field_name = 13

• repeated common.HighlightResult highlight_results = 14

• optional LongArray element_indices = 15

• repeated FieldData group_by_field_values = 17

• repeated AggBucket agg_buckets = 18

• repeated int64 agg_topks = 19

beta, api may change

Used in: VectorField

• repeated bytes contents = 1

• int64 dim = 2

  dim is the max dimension of the current batch of vectors

Used in: IDs, ScalarField, TemplateArrayValue

• repeated string data = 1

Used in: FieldData

• repeated FieldData fields = 1

Used in: milvus.AddCollectionStructFieldRequest, CollectionSchema

• int64 fieldID = 1

• string name = 2

• string description = 3

• repeated FieldSchema fields = 4

• repeated common.KeyValuePair type_params = 5

• bool nullable = 6

Used in: TemplateArrayValueArray, TemplateValue

• oneof data

  • BoolArray bool_data = 1

  • LongArray long_data = 2

  • DoubleArray double_data = 3

  • StringArray string_data = 4

  • TemplateArrayValueArray array_data = 5

  • JSONArray json_data = 6

Used in: TemplateArrayValue

• repeated TemplateArrayValue data = 1

Used in: milvus.DeleteRequest, milvus.QueryRequest, milvus.SearchRequest, milvus.SubSearchRequest

• oneof val

  • bool bool_val = 1

  • int64 int64_val = 2

  • double float_val = 3

  • string string_val = 4

  • TemplateArrayValue array_val = 5

Used in: ScalarField

• repeated int64 data = 1

Used in: FieldSchema

• oneof data

  • bool bool_data = 1

  • int32 int_data = 2

  • int64 long_data = 3

  • float float_data = 4

  • double double_data = 5

  • string string_data = 6

  • bytes bytes_data = 7

  • int64 timestamptz_data = 8

Used in: VectorField

• int64 dim = 1

• repeated VectorField data = 2

• DataType element_type = 3

vector field clustering info

Used in: ClusteringInfo

• string field = 1

  for multi vectors

• optional VectorField centroid = 2

Used in: milvus.VectorsArray, FieldData, VectorArray, VectorClusteringInfo

• int64 dim = 1

  For sparse vector, dim is the max dimension of the current batch of vectors

• oneof data

  • FloatArray float_vector = 2

  • bytes binary_vector = 3

  • bytes float16_vector = 4

  • bytes bfloat16_vector = 5

  • SparseFloatArray sparse_float_vector = 6

  • bytes int8_vector = 7

  • VectorArray vector_array = 8