MCP Tools Reference: vectorsearch.googleapis.com

Tool: search_data_objects

Performs a vector search on a collection.

The following code sample shows how to use curl to call the search_data_objects MCP tool.

Curl Request 
curl --location 'https://vectorsearch.googleapis.com/mcp' \
--header 'content-type: application/json' \
--header 'accept: application/json, text/event-stream' \
--data '{
  "method": "tools/call",
  "params": {
    "name": "search_data_objects",
    "arguments": {
      // Provide these details according to the MCP tool specification.
    }
  },
  "jsonrpc": "2.0",
  "id": 1
}'

Input Schema

Request for performing a single search.

SearchDataObjectsRequest

JSON representation 
{
  "parent": string,
  "pageSize": integer,
  "pageToken": string,

  // Union field search_type can be only one of the following:
  "vectorSearch": {
    object (VectorSearch)
  },
  "semanticSearch": {
    object (SemanticSearch)
  },
  "textSearch": {
    object (TextSearch)
  }
  // End of list of possible types for union field search_type.
}
Fields
parent

string

Required. The resource name of the Collection for which to search. Format: projects/{project}/locations/{location}/collections/{collection}
pageSize

integer

Optional. The standard list page size. Only supported for KNN. If not set, up to search_type.top_k results will be returned. The maximum value is 1000; values above 1000 will be coerced to 1000.
pageToken

string

Optional. The standard list page token. Typically obtained via SearchDataObjectsResponse.next_page_token of the previous DataObjectSearchService.SearchDataObjects call.
Union field search_type. The query to search for. search_type can be only one of the following:
vectorSearch

object (VectorSearch)

A vector search operation.
semanticSearch

object (SemanticSearch)

A semantic search operation.
textSearch

object (TextSearch)

Optional. A text search operation.

VectorSearch

JSON representation 
{
  "searchField": string,
  "filter": {
    object
  },
  "outputFields": {
    object (OutputFields)
  },
  "searchHint": {
    object (SearchHint)
  },
  "distanceMetric": enum (DistanceMetric),

  // Union field vector_type can be only one of the following:
  "vector": {
    object (DenseVector)
  },
  "sparseVector": {
    object (SparseVector)
  }
  // End of list of possible types for union field vector_type.

  // Union field _top_k can be only one of the following:
  "topK": integer
  // End of list of possible types for union field _top_k.
}
Fields
searchField

string

Required. The vector field to search.
filter

object (Struct format)

Optional. A JSON filter expression, e.g. {"genre": {"$eq": "sci-fi"}}, represented as a google.protobuf.Struct.

An empty filter ({}) matches all data objects. Otherwise, each JSON object in the filter must have exactly one key: either a field name mapped to one operator ({"price": {"$lt": 100}}), or a logical operator: $and or $or with a list of filters, or $not with one filter. Field operators and their values: $eq and $ne take a string, number or bool; $lt, $lte, $gt and $gte take a number; $in, $nin and $all take a list of strings; $exists takes a bool. Several keys in one object (implicit AND) and implicit equality ({"genre": "sci-fi"}) are rejected; use $and and $eq instead.

Logical operators can be nested at most 2 levels deep, e.g. {"$and": [{"$or": [{"a": {"$eq": 1}}, {"b": {"$eq": 2}}]}]}. Deeper filters are rejected with INVALID_ARGUMENT.

Which filters are accepted depends on how the search is served. If it is served by an approximate nearest neighbor (ANN) index (by default when a ready index exists on search_field, or when one is selected via search_hint.index_hint), every filtered field must be listed in Index.filter_fields at index creation, $or and $all are accepted only at the top level of the filter, and $not and $exists are not supported. Not every unsupported filter is rejected: most produce INVALID_ARGUMENT, but some are silently ignored, so do not rely on the error. If no ready index exists on search_field, or if search_hint.knn_hint is set, the search is served by K-Nearest Neighbor (KNN) search, which has none of these filter restrictions.
outputFields

object (OutputFields)

Optional. Mask specifying which fields to return.
searchHint

object (SearchHint)

Optional. Sets the search hint. If no strategy is specified, the service will use an index if one is available, and fall back to the default KNN search otherwise.
distanceMetric

enum (DistanceMetric)

Optional. The distance metric to use for the KNN search. If not specified, DOT_PRODUCT will be used as the default.
Union field vector_type. Specifies the type of vector to use for the query. vector_type can be only one of the following:
vector

object (DenseVector)

A dense vector for the query.
sparseVector

object (SparseVector)

A sparse vector for the query.

Union field _top_k.

_top_k can be only one of the following:
topK

integer

Optional. The number of nearest neighbors to return.

DenseVector

JSON representation 
{
  "values": [
    number
  ]
}
Fields
values[]

number

Required. The values of the vector.

SparseVector

JSON representation 
{
  "values": [
    number
  ],
  "indices": [
    integer
  ]
}
Fields
values[]

number

Required. The values of the vector.
indices[]

integer

Required. The corresponding indices for the values.

Struct

JSON representation 
{
  "fields": {
    string: value,
    ...
  }
}
Fields
fields

map (key: string, value: value (Value format))

Unordered map of dynamically typed values.

An object containing a list of "key": value pairs. Example: { "name": "wrench", "mass": "1.3kg", "count": "3" }.

FieldsEntry

JSON representation 
{
  "key": string,
  "value": value
}
Fields
key

string
value

value (Value format)

Value

JSON representation 
{

  // Union field kind can be only one of the following:
  "nullValue": null,
  "numberValue": number,
  "stringValue": string,
  "boolValue": boolean,
  "structValue": {
    object
  },
  "listValue": array
  // End of list of possible types for union field kind.
}
Fields
Union field kind. The kind of value. kind can be only one of the following:
nullValue

null

Represents a JSON null.
numberValue

number

Represents a JSON number. Must not be NaN, Infinity or -Infinity, since those are not supported in JSON. This also cannot represent large Int64 values, since JSON format generally does not support them in its number type.
stringValue

string

Represents a JSON string.
boolValue

boolean

Represents a JSON boolean (true or false literal in JSON).
structValue

object (Struct format)

Represents a JSON object.
listValue

array (ListValue format)

Represents a JSON array.

ListValue

JSON representation 
{
  "values": [
    value
  ]
}
Fields
values[]

value (Value format)

Repeated field of dynamically typed values.

OutputFields

JSON representation 
{
  "dataFields": [
    string
  ],
  "vectorFields": [
    string
  ],
  "metadataFields": [
    string
  ]
}
Fields
dataFields[]

string

Optional. The fields from the data fields to include in the output. Use "*" to include all data fields. If empty, data is omitted from results.
vectorFields[]

string

Optional. The fields from the vector fields to include in the output. Use "*" to include all vector fields. If empty, vectors is omitted from results.
metadataFields[]

string

Optional. The DataObject metadata fields to include in the output: create_time, update_time and etag. Use "*" to include all metadata fields. If empty, these fields are omitted from results.

SearchHint

JSON representation 
{

  // Union field index_type can be only one of the following:
  "knnHint": {
    object (KnnHint)
  },
  "indexHint": {
    object (IndexHint)
  }
  // End of list of possible types for union field index_type.
}
Fields
Union field index_type. The type of index to use. index_type can be only one of the following:
knnHint

object (KnnHint)

Optional. If set, the search will use the system's default K-Nearest Neighbor (KNN) index engine.
indexHint

object (IndexHint)

Optional. Specifies that the search should use a particular index.

IndexHint

JSON representation 
{
  "name": string,

  // Union field params can be only one of the following:
  "denseScannParams": {
    object (DenseScannParams)
  }
  // End of list of possible types for union field params.
}
Fields
name

string

Required. The resource name of the index to use for the search. The index must be in the same project, location, and collection. Format: projects/{project}/locations/{location}/collections/{collection}/indexes/{index}
Union field params. The parameters for the index. params can be only one of the following:
denseScannParams

object (DenseScannParams)

Optional. Dense ScaNN parameters.

DenseScannParams

JSON representation 
{

  // Union field _target_recall can be only one of the following:
  "targetRecall": number
  // End of list of possible types for union field _target_recall.
}
Fields

Union field _target_recall.

_target_recall can be only one of the following:
targetRecall

number

Optional. The target recall for the search. Must be a double in the range [0, 1]. While the search aims to achieve this level of recall, it is not guaranteed.

SemanticSearch

JSON representation 
{
  "searchText": string,
  "searchField": string,
  "taskType": enum (EmbeddingTaskType),
  "outputFields": {
    object (OutputFields)
  },
  "filter": {
    object
  },
  "searchHint": {
    object (SearchHint)
  },

  // Union field _top_k can be only one of the following:
  "topK": integer
  // End of list of possible types for union field _top_k.
}
Fields
searchText

string

Optional. The query text, which is used to generate an embedding according to the embedding model specified in the collection config.

Required when using the text search mode.
searchField

string

Required. The vector field to search.
taskType

enum (EmbeddingTaskType)

Optional. The task type of the query embedding. Must be specified for text-only embedding models, see https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/embeddings/task-types Not needed for multi modal embedding models, see https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/embeddings/get-multimodal-embeddings#specify-task-instructions
outputFields

object (OutputFields)

Optional. The fields to return in the search results.
filter

object (Struct format)

Optional. A JSON filter expression, e.g. {"genre": {"$eq": "sci-fi"}}, represented as a google.protobuf.Struct.

An empty filter ({}) matches all data objects. Otherwise, each JSON object in the filter must have exactly one key: either a field name mapped to one operator ({"price": {"$lt": 100}}), or a logical operator: $and or $or with a list of filters, or $not with one filter. Field operators and their values: $eq and $ne take a string, number or bool; $lt, $lte, $gt and $gte take a number; $in, $nin and $all take a list of strings; $exists takes a bool. Several keys in one object (implicit AND) and implicit equality ({"genre": "sci-fi"}) are rejected; use $and and $eq instead.

Logical operators can be nested at most 2 levels deep, e.g. {"$and": [{"$or": [{"a": {"$eq": 1}}, {"b": {"$eq": 2}}]}]}. Deeper filters are rejected with INVALID_ARGUMENT.

Which filters are accepted depends on how the search is served. If it is served by an approximate nearest neighbor (ANN) index (by default when a ready index exists on search_field, or when one is selected via search_hint.index_hint), every filtered field must be listed in Index.filter_fields at index creation, $or and $all are accepted only at the top level of the filter, and $not and $exists are not supported. Not every unsupported filter is rejected: most produce INVALID_ARGUMENT, but some are silently ignored, so do not rely on the error. If no ready index exists on search_field, or if search_hint.knn_hint is set, the search is served by K-Nearest Neighbor (KNN) search, which has none of these filter restrictions.
searchHint

object (SearchHint)

Optional. Sets the search hint. If no strategy is specified, the service will use an index if one is available, and fall back to KNN search otherwise.

Union field _top_k.

_top_k can be only one of the following:
topK

integer

Optional. The number of data objects to return.

TextSearch

JSON representation 
{
  "searchText": string,
  "dataFieldNames": [
    string
  ],
  "outputFields": {
    object (OutputFields)
  },
  "filter": {
    object
  },

  // Union field _top_k can be only one of the following:
  "topK": integer
  // End of list of possible types for union field _top_k.
}
Fields
searchText

string

Optional. The query text. Required when using the default text search mode.
dataFieldNames[]

string

Optional. The data field names to search. Required when using the default text search mode.
outputFields

object (OutputFields)

Optional. The fields to return in the search results.
filter

object (Struct format)

Optional. A JSON filter expression, e.g. {"genre": {"$eq": "sci-fi"}}, represented as a google.protobuf.Struct.

An empty filter ({}) matches all data objects. Otherwise, each JSON object in the filter must have exactly one key: either a field name mapped to one operator ({"price": {"$lt": 100}}), or a logical operator: $and or $or with a list of filters, or $not with one filter. Field operators and their values: $eq and $ne take a string, number or bool; $lt, $lte, $gt and $gte take a number; $in, $nin and $all take a list of strings; $exists takes a bool. Several keys in one object (implicit AND) and implicit equality ({"genre": "sci-fi"}) are rejected; use $and and $eq instead.

Logical operators can be nested at most 2 levels deep, e.g. {"$and": [{"$or": [{"a": {"$eq": 1}}, {"b": {"$eq": 2}}]}]}. Deeper filters are rejected with INVALID_ARGUMENT.

Union field _top_k.

_top_k can be only one of the following:
topK

integer

Optional. The number of results to return.

NullValue

Represents a JSON null.

NullValue is a sentinel, using an enum with only one value to represent the null value for the Value type union.

A field of type NullValue with any value other than 0 is considered invalid. Most ProtoJSON serializers will emit a Value with a null_value set as a JSON null regardless of the integer value, and so will round trip to a 0 value.

Enums
NULL_VALUE Null value.

DistanceMetric

Distance metric for vector search.

Enums
DISTANCE_METRIC_UNSPECIFIED Default value, distance metric is not specified.
DOT_PRODUCT Dot product distance metric.
COSINE_DISTANCE Cosine distance metric.

EmbeddingTaskType

Represents the task the embeddings will be used for.

Enums
EMBEDDING_TASK_TYPE_UNSPECIFIED Unspecified task type.
RETRIEVAL_QUERY Specifies the given text is a query in a search/retrieval setting.
RETRIEVAL_DOCUMENT Specifies the given text is a document from the corpus being searched.
SEMANTIC_SIMILARITY Specifies the given text will be used for STS.
CLASSIFICATION Specifies that the given text will be classified.
CLUSTERING Specifies that the embeddings will be used for clustering.
QUESTION_ANSWERING Specifies that the embeddings will be used for question answering.
FACT_VERIFICATION Specifies that the embeddings will be used for fact verification.
CODE_RETRIEVAL_QUERY Specifies that the embeddings will be used for code retrieval.

Output Schema

Response for a search request.

SearchDataObjectsResponse

JSON representation 
{
  "results": [
    {
      object (SearchResult)
    }
  ],
  "nextPageToken": string
}
Fields
results[]

object (SearchResult)

Output only. The list of dataObjects that match the search criteria.
nextPageToken

string

Output only. A token to retrieve next page of results. Pass to pageToken to obtain that page.

SearchResult

JSON representation 
{
  "dataObject": {
    object (DataObject)
  },

  // Union field _distance can be only one of the following:
  "distance": number
  // End of list of possible types for union field _distance.
}
Fields
dataObject

object (DataObject)

Output only. The matching data object.

Union field _distance.

_distance can be only one of the following:
distance

number

Output only. Similarity distance or ranker score returned by BatchSearchDataObjects.

DataObject

JSON representation 
{
  "name": string,
  "dataObjectId": string,
  "createTime": string,
  "updateTime": string,
  "data": {
    object
  },
  "vectors": {
    string: {
      object (Vector)
    },
    ...
  },
  "etag": string
}
Fields
name

string

Identifier. The fully qualified resource name of the dataObject.

Format: projects/{project}/locations/{location}/collections/{collection}/dataObjects/{data_object_id} The data_object_id must be 1-63 characters long, and comply with RFC1035.
dataObjectId

string

Output only. The id of the dataObject.
createTime

string (Timestamp format)

Output only. Timestamp the dataObject was created at.

Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z", "2014-10-02T15:01:23.045123456Z" or "2014-10-02T15:01:23+05:30".
updateTime

string (Timestamp format)

Output only. Timestamp the dataObject was last updated.

Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z", "2014-10-02T15:01:23.045123456Z" or "2014-10-02T15:01:23+05:30".
data

object (Struct format)

Optional. The data of the dataObject.
vectors

map (key: string, value: object (Vector))

Optional. The vectors of the dataObject.

An object containing a list of "key": value pairs. Example: { "name": "wrench", "mass": "1.3kg", "count": "3" }.
etag

string

Optional. The etag of the dataObject.

Timestamp

JSON representation 
{
  "seconds": string,
  "nanos": integer
}
Fields
seconds

string (int64 format)

Represents seconds of UTC time since Unix epoch 1970-01-01T00:00:00Z. Must be between -62135596800 and 253402300799 inclusive (which corresponds to 0001-01-01T00:00:00Z to 9999-12-31T23:59:59Z).
nanos

integer

Non-negative fractions of a second at nanosecond resolution. This field is the nanosecond portion of the duration, not an alternative to seconds. Negative second values with fractions must still have non-negative nanos values that count forward in time. Must be between 0 and 999,999,999 inclusive.

Struct

JSON representation 
{
  "fields": {
    string: value,
    ...
  }
}
Fields
fields

map (key: string, value: value (Value format))

Unordered map of dynamically typed values.

An object containing a list of "key": value pairs. Example: { "name": "wrench", "mass": "1.3kg", "count": "3" }.

FieldsEntry

JSON representation 
{
  "key": string,
  "value": value
}
Fields
key

string
value

value (Value format)

Value

JSON representation 
{

  // Union field kind can be only one of the following:
  "nullValue": null,
  "numberValue": number,
  "stringValue": string,
  "boolValue": boolean,
  "structValue": {
    object
  },
  "listValue": array
  // End of list of possible types for union field kind.
}
Fields
Union field kind. The kind of value. kind can be only one of the following:
nullValue

null

Represents a JSON null.
numberValue

number

Represents a JSON number. Must not be NaN, Infinity or -Infinity, since those are not supported in JSON. This also cannot represent large Int64 values, since JSON format generally does not support them in its number type.
stringValue

string

Represents a JSON string.
boolValue

boolean

Represents a JSON boolean (true or false literal in JSON).
structValue

object (Struct format)

Represents a JSON object.
listValue

array (ListValue format)

Represents a JSON array.

ListValue

JSON representation 
{
  "values": [
    value
  ]
}
Fields
values[]

value (Value format)

Repeated field of dynamically typed values.

VectorsEntry

JSON representation 
{
  "key": string,
  "value": {
    object (Vector)
  }
}
Fields
key

string
value

object (Vector)

Vector

JSON representation 
{

  // Union field vector_type can be only one of the following:
  "dense": {
    object (DenseVector)
  },
  "sparse": {
    object (SparseVector)
  }
  // End of list of possible types for union field vector_type.
}
Fields
Union field vector_type. The type of the vector. vector_type can be only one of the following:
dense

object (DenseVector)

A dense vector.
sparse

object (SparseVector)

A sparse vector.

DenseVector

JSON representation 
{
  "values": [
    number
  ]
}
Fields
values[]

number

Required. The values of the vector.

SparseVector

JSON representation 
{
  "values": [
    number
  ],
  "indices": [
    integer
  ]
}
Fields
values[]

number

Required. The values of the vector.
indices[]

integer

Required. The corresponding indices for the values.

NullValue

Represents a JSON null.

NullValue is a sentinel, using an enum with only one value to represent the null value for the Value type union.

A field of type NullValue with any value other than 0 is considered invalid. Most ProtoJSON serializers will emit a Value with a null_value set as a JSON null regardless of the integer value, and so will round trip to a 0 value.

Enums
NULL_VALUE Null value.

Tool Annotations

Tool annotations are sent to MCP clients to describe the basic risk of a given tool. Most clients treat these hints as untrusted, but they can be used to decide when a confirmation prompt might be sent to a user.

Along with the title string, the following boolean hints are defined as follows:

  • readOnlyHint: If true, the tool doesn't modify its environment. Default: false.
  • destructiveHint: If true, then the tool can perform destructive actions. If false, then the tool can only perform additive actions. Default: true.
  • idempotentHint: If true, then calling the tool repeatedly with the same arguments will have no additional effect on its environment. Default: false.
  • openWorldHint: If true, then the tool can interact with an 'open world' of external entities. If false, then the tool can only interact with internal entities. For example, a web search tool would be open world, while a memory tool would not be open world.

Destructive Hint: ❌ | Idempotent Hint: ✅ | Read Only Hint: ✅ | Open World Hint: ❌