REST Resource: projects.locations.collections.dataStores.servingConfigs

Resource: ServingConfig

Configures metadata that is used to generate serving time results (e.g. search results or recommendation predictions). The ServingConfig is passed in the search and predict request and generates results.

JSON representation 
{
  "name": string,
  "displayName": string,
  "solutionType": enum (SolutionType),
  "modelId": string,
  "diversityLevel": string,
  "rankingExpression": string,
  "createTime": string,
  "updateTime": string,
  "filterControlIds": [
    string
  ],
  "boostControlIds": [
    string
  ],
  "redirectControlIds": [
    string
  ],
  "synonymsControlIds": [
    string
  ],
  "onewaySynonymsControlIds": [
    string
  ],
  "dissociateControlIds": [
    string
  ],
  "replacementControlIds": [
    string
  ],
  "ignoreControlIds": [
    string
  ],
  "promoteControlIds": [
    string
  ],
  "answerGenerationSpec": {
    object (AnswerGenerationSpec)
  },

  // Union field vertical_config can be only one of the following:
  "mediaConfig": {
    object (MediaConfig)
  },
  "genericConfig": {
    object (GenericConfig)
  }
  // End of list of possible types for union field vertical_config.
}
Fields
name

string

Immutable. Fully qualified name projects/{project}/locations/{location}/collections/{collectionId}/engines/{engineId}/servingConfigs/{servingConfigId}
displayName

string

Required. The human readable serving config display name. Used in Discovery UI.

This field must be a UTF-8 encoded string with a length limit of 128 characters. Otherwise, an INVALID_ARGUMENT error is returned.
solutionType

enum (SolutionType)

Required. Immutable. Specifies the solution type that a serving config can be associated with.
modelId

string

The ID of the model to use at serving time. Currently only RecommendationModels are supported. Can be changed but only to a compatible model (e.g. others-you-may-like CTR to others-you-may-like CVR).

Required when SolutionType is SOLUTION_TYPE_RECOMMENDATION.
diversityLevel

string

How much diversity to use in recommendation model results e.g. medium-diversity or high-diversity. Currently supported values:

  • no-diversity
  • low-diversity
  • medium-diversity
  • high-diversity
  • auto-diversity

If not specified, we choose default based on recommendation model type. Default value: no-diversity.

Can only be set if SolutionType is SOLUTION_TYPE_RECOMMENDATION.
rankingExpression

string

The ranking expression controls the customized ranking on retrieval documents. To leverage this, document embedding is required. The ranking expression setting in ServingConfig applies to all search requests served by the serving config. However, if SearchRequest.ranking_expression is specified, it overrides the ServingConfig ranking expression.

The ranking expression is a single function or multiple functions that are joined by "+".

  • rankingExpression = function, { " + ", function };

Supported functions:

  • double * relevanceScore
  • double * dotProduct(embedding_field_path)

Function variables:

  • relevanceScore: pre-defined keywords, used for measure relevance between query and document.
  • embedding_field_path: the document embedding field used with query embedding vector.
  • dotProduct: embedding function between embedding_field_path and query embedding vector.

Example ranking expression:

If document has an embedding field doc_embedding, the ranking expression could be 0.5 * relevanceScore + 0.3 * dotProduct(doc_embedding).
createTime

string (Timestamp format)

Output only. ServingConfig created timestamp.

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. ServingConfig updated timestamp.

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".
filterControlIds[]

string

Filter controls to use in serving path. All triggered filter controls will be applied. Filter controls must be in the same data store as the serving config. Maximum of 20 filter controls.
boostControlIds[]

string

Boost controls to use in serving path. All triggered boost controls will be applied. Boost controls must be in the same data store as the serving config. Maximum of 20 boost controls.
redirectControlIds[]

string

IDs of the redirect controls. Only the first triggered redirect action is applied, even if multiple apply. Maximum number of specifications is 100.

Can only be set if SolutionType is SOLUTION_TYPE_SEARCH.
synonymsControlIds[]

string

Condition synonyms specifications. If multiple synonyms conditions match, all matching synonyms controls in the list will execute. Maximum number of specifications is 100.

Can only be set if SolutionType is SOLUTION_TYPE_SEARCH.
onewaySynonymsControlIds[]

string

Condition oneway synonyms specifications. If multiple oneway synonyms conditions match, all matching oneway synonyms controls in the list will execute. Maximum number of specifications is 100.

Can only be set if SolutionType is SOLUTION_TYPE_SEARCH.
dissociateControlIds[]

string

Condition do not associate specifications. If multiple do not associate conditions match, all matching do not associate controls in the list will execute. Order does not matter. Maximum number of specifications is 100.

Can only be set if SolutionType is SOLUTION_TYPE_SEARCH.
replacementControlIds[]

string

Condition replacement specifications. Applied according to the order in the list. A previously replaced term can not be re-replaced. Maximum number of specifications is 100.

Can only be set if SolutionType is SOLUTION_TYPE_SEARCH.
ignoreControlIds[]

string

Condition ignore specifications. If multiple ignore conditions match, all matching ignore controls in the list will execute. Order does not matter. Maximum number of specifications is 100.
promoteControlIds[]

string

Condition promote specifications.

Maximum number of specifications is 100.
answerGenerationSpec

object (AnswerGenerationSpec)

Optional. The specification for answer generation.
Union field vertical_config. Industry vertical specific config. vertical_config can be only one of the following:
mediaConfig

object (MediaConfig)

The MediaConfig of the serving configuration.
genericConfig

object (GenericConfig)

The GenericConfig of the serving configuration.

MediaConfig

Specifies the configurations needed for Media Discovery. Currently we support:

  • demote_content_watched: Threshold for watched content demotion. Customers can specify if using watched content demotion or use viewed detail page. Using the content watched demotion, customers need to specify the watched minutes or percentage exceeds the threshold, the content will be demoted in the recommendation result.
  • promote_fresh_content: cutoff days for fresh content promotion. Customers can specify if using content freshness promotion. If the content was published within the cutoff days, the content will be promoted in the recommendation result. Can only be set if SolutionType is SOLUTION_TYPE_RECOMMENDATION.
JSON representation 
{
  "demotionEventType": string,
  "demoteContentWatchedPastDays": integer,
  "contentFreshnessCutoffDays": integer,

  // Union field demote_content_watched can be only one of the following:
  "contentWatchedPercentageThreshold": number,
  "contentWatchedSecondsThreshold": number
  // End of list of possible types for union field demote_content_watched.
}
Fields
demotionEventType

string

Specifies the event type used for demoting recommendation result. Currently supported values:

  • view-item: Item viewed.
  • media-play: Start/resume watching a video, playing a song, etc.
  • media-complete: Finished or stopped midway through a video, song, etc.

If unset, watch history demotion will not be applied. Content freshness demotion will still be applied.
demoteContentWatchedPastDays

integer

Optional. Specifies the number of days to look back for demoting watched content. If set to zero or unset, defaults to the maximum of 365 days.
contentFreshnessCutoffDays

integer

Specifies the content freshness used for recommendation result. Contents will be demoted if contents were published for more than content freshness cutoff days.
Union field demote_content_watched. Specify the threshold for demoting watched content, the threshold can be either percentage or minutes value. This must be set for media-complete event type. demote_content_watched can be only one of the following:
contentWatchedPercentageThreshold

number

Specifies the content watched percentage threshold for demotion. Threshold value must be between [0, 1.0] inclusive.
contentWatchedSecondsThreshold

number

Specifies the content watched minutes threshold for demotion.

GenericConfig

Specifies the configurations needed for Generic Discovery.Currently we support:

  • contentSearchSpec: configuration for generic content search.
JSON representation 
{
  "contentSearchSpec": {
    object (ContentSearchSpec)
  }
}
Fields
contentSearchSpec

object (ContentSearchSpec)

Specifies the expected behavior of content search. Only valid for content-search enabled data store.

ContentSearchSpec

A specification for configuring the behavior of content search.

JSON representation 
{
  "snippetSpec": {
    object (SnippetSpec)
  },
  "summarySpec": {
    object (SummarySpec)
  },
  "extractiveContentSpec": {
    object (ExtractiveContentSpec)
  },
  "searchResultMode": enum (SearchResultMode),
  "chunkSpec": {
    object (ChunkSpec)
  }
}
Fields
snippetSpec

object (SnippetSpec)

If snippetSpec is not specified, snippets are not included in the search response.
summarySpec

object (SummarySpec)

If summarySpec is not specified, summaries are not included in the search response.
extractiveContentSpec

object (ExtractiveContentSpec)

If there is no extractiveContentSpec provided, there will be no extractive answer in the search response.
searchResultMode

enum (SearchResultMode)

Specifies the search result mode. If unspecified, the search result mode defaults to DOCUMENTS.
chunkSpec

object (ChunkSpec)

Specifies the chunk spec to be returned from the search response. Only available if the SearchRequest.ContentSearchSpec.search_result_mode is set to CHUNKS

SnippetSpec

A specification for configuring snippets in a search response.

JSON representation 
{
  "maxSnippetCount": integer,
  "referenceOnly": boolean,
  "returnSnippet": boolean
}
Fields
maxSnippetCount
(deprecated)

integer

[DEPRECATED] This field is deprecated. To control snippet return, use returnSnippet field. For backwards compatibility, we will return snippet if maxSnippetCount > 0.
referenceOnly
(deprecated)

boolean

[DEPRECATED] This field is deprecated and will have no affect on the snippet.
returnSnippet

boolean

If true, then return snippet. If no snippet can be generated, we return "No snippet is available for this page." A snippetStatus with SUCCESS or NO_SNIPPET_AVAILABLE will also be returned.

SummarySpec

A specification for configuring a summary returned in a search response.

JSON representation 
{
  "summaryResultCount": integer,
  "includeCitations": boolean,
  "ignoreAdversarialQuery": boolean,
  "ignoreNonSummarySeekingQuery": boolean,
  "ignoreLowRelevantContent": boolean,
  "ignoreJailBreakingQuery": boolean,
  "modelPromptSpec": {
    object (ModelPromptSpec)
  },
  "languageCode": string,
  "modelSpec": {
    object (ModelSpec)
  },
  "useSemanticChunks": boolean
}
Fields
summaryResultCount

integer

The number of top results to generate the summary from. If the number of results returned is less than summaryResultCount, the summary is generated from all of the results.

At most 10 results for documents mode, or 50 for chunks mode, can be used to generate a summary. The chunks mode is used when SearchRequest.ContentSearchSpec.search_result_mode is set to CHUNKS.
includeCitations

boolean

Specifies whether to include citations in the summary. The default value is false.

When this field is set to true, summaries include in-line citation numbers.

Example summary including citations:

BigQuery is Google Cloud's fully managed and completely serverless enterprise data warehouse [1]. BigQuery supports all data types, works across clouds, and has built-in machine learning and business intelligence, all within a unified platform [2, 3].

The citation numbers refer to the returned search results and are 1-indexed. For example, [1] means that the sentence is attributed to the first search result. [2, 3] means that the sentence is attributed to both the second and third search results.
ignoreAdversarialQuery

boolean

Specifies whether to filter out adversarial queries. The default value is false.

Google employs search-query classification to detect adversarial queries. No summary is returned if the search query is classified as an adversarial query. For example, a user might ask a question regarding negative comments about the company or submit a query designed to generate unsafe, policy-violating output. If this field is set to true, we skip generating summaries for adversarial queries and return fallback messages instead.
ignoreNonSummarySeekingQuery

boolean

Specifies whether to filter out queries that are not summary-seeking. The default value is false.

Google employs search-query classification to detect summary-seeking queries. No summary is returned if the search query is classified as a non-summary seeking query. For example, why is the sky blue and Who is the best soccer player in the world? are summary-seeking queries, but SFO airport and world cup 2026 are not. They are most likely navigational queries. If this field is set to true, we skip generating summaries for non-summary seeking queries and return fallback messages instead.
ignoreLowRelevantContent

boolean

Specifies whether to filter out queries that have low relevance. The default value is false.

If this field is set to false, all search results are used regardless of relevance to generate answers. If set to true, only queries with high relevance search results will generate answers.
ignoreJailBreakingQuery

boolean

Optional. Specifies whether to filter out jail-breaking queries. The default value is false.

Google employs search-query classification to detect jail-breaking queries. No summary is returned if the search query is classified as a jail-breaking query. A user might add instructions to the query to change the tone, style, language, content of the answer, or ask the model to act as a different entity, e.g. "Reply in the tone of a competing company's CEO". If this field is set to true, we skip generating summaries for jail-breaking queries and return fallback messages instead.
modelPromptSpec

object (ModelPromptSpec)

If specified, the spec will be used to modify the prompt provided to the LLM.
languageCode

string

Language code for Summary. Use language tags defined by BCP47. Note: This is an experimental feature.
modelSpec

object (ModelSpec)

If specified, the spec will be used to modify the model specification provided to the LLM.
useSemanticChunks

boolean

If true, answer will be generated from most relevant chunks from top search results. This feature will improve summary quality. Note that with this feature enabled, not all top search results will be referenced and included in the reference list, so the citation source index only points to the search results listed in the reference list.

ModelPromptSpec

Specification of the prompt to use with the model.

JSON representation 
{
  "preamble": string
}
Fields
preamble

string

Text at the beginning of the prompt that instructs the assistant. Examples are available in the user guide.

ModelSpec

Specification of the model.

JSON representation 
{
  "version": string
}
Fields
version

string

The model version used to generate the summary.

Supported values are:

ExtractiveContentSpec

A specification for configuring the extractive content in a search response.

JSON representation 
{
  "maxExtractiveAnswerCount": integer,
  "maxExtractiveSegmentCount": integer,
  "returnExtractiveSegmentScore": boolean,
  "numPreviousSegments": integer,
  "numNextSegments": integer
}
Fields
maxExtractiveAnswerCount

integer

The maximum number of extractive answers returned in each search result.

An extractive answer is a verbatim answer extracted from the original document, which provides a precise and contextually relevant answer to the search query.

If the number of matching answers is less than the maxExtractiveAnswerCount, return all of the answers. Otherwise, return the maxExtractiveAnswerCount.

At most five answers are returned for each SearchResult.
maxExtractiveSegmentCount

integer

The max number of extractive segments returned in each search result. Only applied if the DataStore is set to DataStore.ContentConfig.CONTENT_REQUIRED or DataStore.solution_types is SOLUTION_TYPE_CHAT.

An extractive segment is a text segment extracted from the original document that is relevant to the search query, and, in general, more verbose than an extractive answer. The segment could then be used as input for LLMs to generate summaries and answers.

If the number of matching segments is less than maxExtractiveSegmentCount, return all of the segments. Otherwise, return the maxExtractiveSegmentCount.
returnExtractiveSegmentScore

boolean

Specifies whether to return the confidence score from the extractive segments in each search result. This feature is available only for new or allowlisted data stores. To allowlist your data store, contact your Customer Engineer. The default value is false.
numPreviousSegments

integer

Specifies whether to also include the adjacent from each selected segments. Return at most numPreviousSegments segments before each selected segments.
numNextSegments

integer

Return at most numNextSegments segments after each selected segments.

SearchResultMode

Specifies the search result mode. If unspecified, the search result mode defaults to DOCUMENTS.

Enums
SEARCH_RESULT_MODE_UNSPECIFIED Default value.
DOCUMENTS Returns documents in the search result.
CHUNKS Returns chunks in the search result. Only available if the DocumentProcessingConfig.chunking_config is specified.

ChunkSpec

Specifies the chunk spec to be returned from the search response. Only available if the SearchRequest.ContentSearchSpec.search_result_mode is set to CHUNKS

JSON representation 
{
  "numPreviousChunks": integer,
  "numNextChunks": integer
}
Fields
numPreviousChunks

integer

The number of previous chunks to be returned of the current chunk. The maximum allowed value is 3. If not specified, no previous chunks will be returned.
numNextChunks

integer

The number of next chunks to be returned of the current chunk. The maximum allowed value is 3. If not specified, no next chunks will be returned.

AnswerGenerationSpec

The specification for answer generation.

JSON representation 
{
  "userDefinedClassifierSpec": {
    object (UserDefinedClassifierSpec)
  }
}
Fields
userDefinedClassifierSpec

object (UserDefinedClassifierSpec)

Optional. The specification for user specified classifier spec.

UserDefinedClassifierSpec

The specification for user defined classifier.

JSON representation 
{
  "enableUserDefinedClassifier": boolean,
  "preamble": string,
  "modelId": string,
  "taskMarker": string,
  "topP": number,
  "topK": string,
  "temperature": number,
  "seed": integer
}
Fields
enableUserDefinedClassifier

boolean

Optional. Whether or not to enable and include user defined classifier.
preamble

string

Optional. The preamble to be used for the user defined classifier.
modelId

string

Optional. The model id to be used for the user defined classifier.
taskMarker

string

Optional. The task marker to be used for the user defined classifier.
topP

number

Optional. The top-p value to be used for the user defined classifier.
topK

string (int64 format)

Optional. The top-k value to be used for the user defined classifier.
temperature

number

Optional. The temperature value to be used for the user defined classifier.
seed

integer

Optional. The seed value to be used for the user defined classifier.

Methods

answer

 Answer query method.

get

 Gets a ServingConfig.

list

 Lists all ServingConfigs linked to this dataStore.

patch

 Updates a ServingConfig.

recommend

 Makes a recommendation, which requires a contextual user event.
Performs a search.

searchLite

 Performs a search.

streamAnswer

 Answer query method (streaming).