Message

A message from an interaction between the user and the system.

JSON representation 
{
  "timestamp": string,
  "messageId": string,

  // Union field kind can be only one of the following:
  "userMessage": {
    object (UserMessage)
  },
  "systemMessage": {
    object (SystemMessage)
  }
  // End of list of possible types for union field kind.
}
Fields
timestamp

string (Timestamp format)

Output only. For user messages, this is the time at which the system received the message. For system messages, this is the time at which the system generated the message.

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".
messageId

string

Optional. unique id of the message in the conversation for persistence.
Union field kind. The kind of message. kind can be only one of the following:
userMessage

object (UserMessage)

A message from the user that is interacting with the system.
systemMessage

object (SystemMessage)

A message from the system in response to the user.

UserMessage

A message from the user that is interacting with the system.

JSON representation 
{

  // Union field kind can be only one of the following:
  "text": string
  // End of list of possible types for union field kind.
}
Fields
Union field kind. The kind of content in the user message. kind can be only one of the following:
text

string

Text should use this field instead of blob.

SystemMessage

A message from the system in response to the user. This message can also be a message from the user as historical context for multi-turn conversations with the system.

JSON representation 
{
  "citation": {
    object (Citation)
  },

  // Union field kind can be only one of the following:
  "text": {
    object (TextMessage)
  },
  "schema": {
    object (SchemaMessage)
  },
  "data": {
    object (DataMessage)
  },
  "analysis": {
    object (AnalysisMessage)
  },
  "chart": {
    object (ChartMessage)
  },
  "error": {
    object (ErrorMessage)
  },
  "exampleQueries": {
    object (ExampleQueries)
  }
  // End of list of possible types for union field kind.
  "groupId": integer
}
Fields
citation

object (Citation)

Output only. Citation information for the system message.
Union field kind. The kind of content in the system message. kind can be only one of the following:
text

object (TextMessage)

A direct natural language response to the user message.
schema

object (SchemaMessage)

A message produced during schema resolution.
data

object (DataMessage)

A message produced during data retrieval.
analysis

object (AnalysisMessage)

A message produced during analysis.
chart

object (ChartMessage)

A message produced during chart generation.
error

object (ErrorMessage)

An error message.
exampleQueries

object (ExampleQueries)

Optional. A message containing example queries.
groupId

integer

Identifies the group that the event belongs to. Similar events are deemed to be logically relevant to each other and should be shown together in the UI.

TextMessage

A multi-part text message.

JSON representation 
{
  "parts": [
    string
  ],
  "textType": enum (TextType),
  "thoughtSignature": string
}
Fields
parts[]

string

Optional. The parts of the message.
textType

enum (TextType)

Optional. The type of the text message.
thoughtSignature

string (bytes format)

Optional. An opaque signature for a thought so it can be reused in subsequent requests.

A base64-encoded string.

TextType

The type of the text message.

Enums
TEXT_TYPE_UNSPECIFIED The default text type.
FINAL_RESPONSE The text is a final response to the user question.
THOUGHT The text is a thought from the model.
PROGRESS The text is an informational message about the agent's progress, such as a tool being invoked. This is distinct from the agent's internal thought process (THOUGHT) and the final answer to the user (FINAL_RESPONSE). These messages provide insight into the agent's actions.
FOLLOWUP_QUESTIONS The text is a list of follow-up questions suggested. Each item in parts is a follow-up question.

SchemaMessage

A message produced during schema resolution.

JSON representation 
{

  // Union field kind can be only one of the following:
  "query": {
    object (SchemaQuery)
  },
  "result": {
    object (SchemaResult)
  }
  // End of list of possible types for union field kind.
}
Fields
Union field kind. Whether this message contains the query or the result of the schema resolution. kind can be only one of the following:
query

object (SchemaQuery)

A schema resolution query.
result

object (SchemaResult)

The result of a schema resolution query.

SchemaQuery

A query for resolving the schema relevant to the posed question.

JSON representation 
{
  "question": string
}
Fields
question

string

Optional. The question to send to the system for schema resolution.

SchemaResult

The result of schema resolution.

JSON representation 
{
  "datasources": [
    {
      object (Datasource)
    }
  ]
}
Fields
datasources[]

object (Datasource)

Optional. The datasources used to resolve the schema query.

Datasource

A datasource that can be used to answer questions.

JSON representation 
{
  "schema": {
    object (Schema)
  },
  "structSchema": {
    object
  },

  // Union field reference can be only one of the following:
  "bigqueryTableReference": {
    object (BigQueryTableReference)
  },
  "studioDatasourceId": string,
  "lookerExploreReference": {
    object (LookerExploreReference)
  },
  "bigqueryPropertyGraphReference": {
    object (BigQueryPropertyGraphReference)
  }
  // End of list of possible types for union field reference.
}
Fields
schema

object (Schema)

Optional. The schema of the datasource.
structSchema

object (Struct format)

Optional. A struct representation of the schema. This is populated for datasources with schemas that cannot be fully represented by the strongly-typed schema field.

For Looker datasources, this maps to the LookmlModelExplore type: https://cloud.google.com/looker/docs/reference/looker-api/latest/types/LookmlModelExplore
Union field reference. The reference to the datasource. reference can be only one of the following:
bigqueryTableReference

object (BigQueryTableReference)

A reference to a BigQuery table.
studioDatasourceId

string

A reference to a Looker Studio datasource.
lookerExploreReference

object (LookerExploreReference)

A reference to a Looker explore.
bigqueryPropertyGraphReference

object (BigQueryPropertyGraphReference)

A reference to a BigQuery property graph.

BigQueryPropertyGraphReference

Message representing a reference to a single BigQuery property graph.

JSON representation 
{
  "projectId": string,
  "datasetId": string,
  "propertyGraphId": string
}
Fields
projectId

string

Required. The project that the property graph belongs to.
datasetId

string

Required. The dataset that the property graph belongs to.
propertyGraphId

string

Required. The property graph id.

DataMessage

A message produced during data retrieval.

JSON representation 
{

  // Union field kind can be only one of the following:
  "query": {
    object (DataQuery)
  },
  "generatedSql": string,
  "result": {
    object (DataResult)
  },
  "bigQueryJob": {
    object (BigQueryJob)
  },
  "matchedQuery": {
    object (MatchedQuery)
  }
  // End of list of possible types for union field kind.
}
Fields
Union field kind. Whether this message contains the query, the result, or generated SQL for the data retrieval. kind can be only one of the following:
query

object (DataQuery)

A data retrieval query.
generatedSql

string

SQL generated by the system to retrieve data.
result

object (DataResult)

Retrieved data.
bigQueryJob

object (BigQueryJob)

A BigQuery job executed by the system to retrieve data.
matchedQuery

object (MatchedQuery)

A pre-existing query that was matched to retrieve data.

DataQuery

A query for retrieving data.

JSON representation 
{
  "question": string,
  "name": string,
  "datasources": [
    {
      object (Datasource)
    }
  ],

  // Union field query_type can be only one of the following:
  "looker": {
    object (LookerQuery)
  }
  // End of list of possible types for union field query_type.
}
Fields
question

string

Optional. A natural language question to answer.
name

string

Optional. A snake-case name for the query that reflects its intent. It is used to name the corresponding data result, so that it can be referenced in later steps.

  • Example: "total_sales_by_product"
  • Example: "sales_for_product_12345"
datasources[]

object (Datasource)

Optional. The datasources available to answer the question.
Union field query_type. The type of query to execute. query_type can be only one of the following:
looker

object (LookerQuery)

Optional. A query for retrieving data from a Looker explore.

DataResult

Retrieved data.

JSON representation 
{
  "name": string,
  "schema": {
    object (Schema)
  },
  "data": [
    {
      object
    }
  ],
  "formattedData": [
    {
      object
    }
  ]
}
Fields
name

string

Optional. A snake-case name for the data result that reflects its contents. The name is used to pass the result around by reference, and serves as a signal about its meaning.

  • Example: "total_sales_by_product"
  • Example: "sales_for_product_12345"
schema

object (Schema)

Optional. The schema of the data.
data[]

object (Struct format)

Optional. The content of the data. Each row is a struct that matches the schema. Simple values are represented as strings, while nested structures are represented as lists or structs.
formattedData[]

object (Struct format)

Optional. Formatted representation of the data, when applicable. Each row is a struct that directly corresponds to the row at the same index within the data field. Its values are string representations of the original data, formatted according to data source specifications (e.g., "$1,234.56" for currency). Columns without formatting will default to their raw value representation. If no columns have formatting rules, this field will be empty.

BigQueryJob

A BigQuery job executed by the system.

JSON representation 
{
  "projectId": string,
  "jobId": string,
  "location": string,
  "destinationTable": {
    object (BigQueryTableReference)
  },
  "schema": {
    object (Schema)
  }
}
Fields
projectId

string

Required. The project that the job belongs to.

See JobReference.
jobId

string

Required. The ID of the job.

See JobReference.
location

string

Optional. The location of the job.

See JobReference.
destinationTable

object (BigQueryTableReference)

Optional. A reference to the destination table of the job's query results.

See JobConfigurationQuery.
schema

object (Schema)

Optional. The schema of the job's query results.

See JobStatistics2.

MatchedQuery

A matched query message represents the agent having matched one of the example queries supplied in context as being applicable to the current question. It will also contain additional info during the matching process.

JSON representation 
{
  "exampleQuery": {
    object (ExampleQuery)
  },
  "queryParameterValues": [
    {
      object (QueryParameterValues)
    }
  ]
}
Fields
exampleQuery

object (ExampleQuery)

The query that was matched based on an example query.
queryParameterValues[]

object (QueryParameterValues)

The extracted values for the query parameters.

QueryParameterValues

A query parameter values message represents the values for the query parameters that were extracted from the user question by LLM, based on the example query.

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

string

Required. The name of the parameter.
value

string

Required. The value of the parameter.

AnalysisMessage

A message produced during analysis.

JSON representation 
{

  // Union field kind can be only one of the following:
  "query": {
    object (AnalysisQuery)
  },
  "progressEvent": {
    object (AnalysisEvent)
  }
  // End of list of possible types for union field kind.
}
Fields
Union field kind. Whether this message contains the query or one of the events from the analysis. kind can be only one of the following:
query

object (AnalysisQuery)

An analysis query.
progressEvent

object (AnalysisEvent)

An event indicating the progress of the analysis.

AnalysisQuery

A query for performing an analysis.

JSON representation 
{
  "question": string,
  "dataResultNames": [
    string
  ]
}
Fields
question

string

Optional. An analysis question to help answer the user's original question.
dataResultNames[]

string

Optional. The names of previously retrieved data results to analyze.

AnalysisEvent

An event indicating the progress of an analysis.

JSON representation 
{

  // Union field kind can be only one of the following:
  "plannerReasoning": string,
  "coderInstruction": string,
  "code": string,
  "executionOutput": string,
  "executionError": string,
  "resultVegaChartJson": string,
  "resultNaturalLanguage": string,
  "resultCsvData": string,
  "resultReferenceData": string,
  "error": string
  // End of list of possible types for union field kind.
}
Fields
Union field kind. The kind of event that occurred during the analysis. kind can be only one of the following:
plannerReasoning

string

Python codegen planner's reasoning.
coderInstruction

string

Instructions issued for code generation.
code

string

Generated code.
executionOutput

string

Output from code execution.
executionError

string

An error from code execution.
resultVegaChartJson

string

Result as Vega chart JSON string.
resultNaturalLanguage

string

Result as NL string.
resultCsvData

string

Result as CSV string.
resultReferenceData

string

Result as a reference to a data source.
error

string

A generic error message.

ChartMessage

A message produced during chart generation.

JSON representation 
{

  // Union field kind can be only one of the following:
  "query": {
    object (ChartQuery)
  },
  "result": {
    object (ChartResult)
  }
  // End of list of possible types for union field kind.
}
Fields
Union field kind. Whether this message contains the query or the result of the chart generation. kind can be only one of the following:
query

object (ChartQuery)

A query for generating a chart.
result

object (ChartResult)

The result of a chart generation query.

ChartQuery

A query for generating a chart.

JSON representation 
{
  "instructions": string,
  "dataResultName": string
}
Fields
instructions

string

Optional. Natural language instructions for generating the chart.
dataResultName

string

Optional. The name of a previously retrieved data result to use in the chart.

ChartResult

The result of a chart generation query.

JSON representation 
{
  "vegaConfig": {
    object
  },
  "image": {
    object (Blob)
  }
}
Fields
vegaConfig

object (Struct format)

Optional. A generated Vega chart config. See https://vega.github.io/vega/docs/config/
image

object (Blob)

Optional. A rendering of the chart if this was requested in the context.

Blob

A blob of data with a MIME type.

JSON representation 
{
  "mimeType": string,
  "data": string
}
Fields
mimeType

string

Required. The IANA standard MIME type of the message data.
data

string (bytes format)

Required. The data represented as bytes.

A base64-encoded string.

ErrorMessage

An error message from a tool call. This message is used to represent an error that occurred while an agent was trying to use a tool. It's important to note that not all errors are terminal. Many are recoverable, and the agent may use the information from this error message to self-correct and retry the tool call or try a different approach.

For example, if a data query fails, the agent might receive an ErrorMessage, analyze it, and then generate a corrected query.

Clients should be cautious about interpreting this message as a definitive failure. It can be part of the agent's normal, iterative process of completing a task. Surfacing these errors directly to end-users without context (e.g., as a "hard failure") may be misleading.

JSON representation 
{
  "text": string
}
Fields
text

string

Output only. The text of the error.

ExampleQueries

A message containing derived and authored example queries.

JSON representation 
{
  "exampleQueries": [
    {
      object (ExampleQuery)
    }
  ]
}
Fields
exampleQueries[]

object (ExampleQuery)

Optional. A list of derived and authored example queries, providing examples of relevant and commonly used SQL queries and their corresponding natural language queries optionally present. Currently only used for BigQuery data sources.

Citation

Source attributions for content.

JSON representation 
{
  "sources": [
    {
      object (CitationSource)
    }
  ],
  "anchors": [
    {
      object (CitationAnchor)
    }
  ]
}
Fields
sources[]

object (CitationSource)

Output only. List of the sources being cited.
anchors[]

object (CitationAnchor)

Output only. List of the anchors of the citations.

CitationSource

The source of the citation.

JSON representation 
{
  "id": string,
  "title": string,

  // Union field source_type can be only one of the following:
  "uri": string,
  "exampleQuery": {
    object (ExampleQuery)
  },
  "glossaryTerm": {
    object (GlossaryTerm)
  }
  // End of list of possible types for union field source_type.
}
Fields
id

string

Output only. Unique identifier of the source. This ID is service-generated and is unique within the scope of a single Citation message.
title

string

Output only. The title of the source.
Union field source_type. The source of the citation, which can be one of the supported types. source_type can be only one of the following:
uri

string

Output only. The uri used as the source, such as a web grounding URL.
exampleQuery

object (ExampleQuery)

Output only. The example query used as the source.
glossaryTerm

object (GlossaryTerm)

Output only. The glossary term used as the source.

CitationAnchor

The anchor of the citation.

JSON representation 
{

  // Union field anchor_type can be only one of the following:
  "textMessageAnchor": {
    object (TextMessageCitationAnchor)
  }
  // End of list of possible types for union field anchor_type.
}
Fields
Union field anchor_type. The anchor of the citation, which can be one of the supported types. anchor_type can be only one of the following:
textMessageAnchor

object (TextMessageCitationAnchor)

Output only. Only set if the citation is for a TextMessage.

TextMessageCitationAnchor

Citation anchor within a TextMessage.

JSON representation 
{
  "partIndex": integer,
  "startOffsetBytes": integer,
  "endOffsetBytes": integer,
  "sourceIds": [
    string
  ]
}
Fields
partIndex

integer

Output only. The 0-based index of the part within the TextMessage.parts field.
startOffsetBytes

integer

Output only. The offset, measured in UTF-8 bytes, within the part string where the citation begins (inclusive). Example: For the text "Hello, world" where "world" is cited, the start offset bytes (inclusive) is 7 and the end offset bytes (exclusive) is 12.
endOffsetBytes

integer

Output only. The offset, measured in UTF-8 bytes, within the part string where the citation ends (exclusive). Example: For the text "Hello, world" where "world" is cited, the start offset bytes (inclusive) is 7 and the end offset bytes (exclusive) is 12.
sourceIds[]

string

Output only. The ids of the sources that are cited.