@google/genai 1.51.0 → 2.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -487,6 +487,30 @@ declare interface AudioContent {
487
487
  uri?: string;
488
488
  }
489
489
 
490
+ /**
491
+ * Configuration for audio output format.
492
+ */
493
+ declare interface AudioResponseFormat {
494
+ type: 'audio';
495
+ /**
496
+ * Bit rate in bits per second (bps). Only applicable for compressed formats (MP3,
497
+ * Opus).
498
+ */
499
+ bitRate?: number;
500
+ /**
501
+ * The delivery mode for the audio output.
502
+ */
503
+ delivery?: 'inline' | 'url';
504
+ /**
505
+ * The MIME type of the audio output.
506
+ */
507
+ mimeType?: 'audio/mp3' | 'audio/ogg_opus' | 'audio/l16' | 'audio/wav' | 'audio/alaw' | 'audio/mulaw';
508
+ /**
509
+ * Sample rate in Hz.
510
+ */
511
+ sampleRate?: number;
512
+ }
513
+
490
514
  /** The audio transcription configuration in Setup. */
491
515
  export declare interface AudioTranscriptionConfig {
492
516
  /** The language codes of the audio. BCP-47 language code. If not set, the transcription will be in the language detected by the model. If set, the server will use the language code specified in the model config as a hint for the language of the audio
@@ -659,7 +683,7 @@ declare interface BaseCreateAgentInteractionParams {
659
683
  /**
660
684
  * Body param: The input for the interaction.
661
685
  */
662
- input: Array<Content_2> | string | Array<Turn> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | CodeExecutionCallContent | URLContextCallContent | MCPServerToolCallContent | GoogleSearchCallContent | FileSearchCallContent | GoogleMapsCallContent | FunctionResultContent | CodeExecutionResultContent | URLContextResultContent | GoogleSearchResultContent | MCPServerToolResultContent | FileSearchResultContent | GoogleMapsResultContent;
686
+ input: string | Array<Step> | Array<Content_2> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent;
663
687
  /**
664
688
  * Body param: Configuration parameters for the agent interaction.
665
689
  */
@@ -676,7 +700,7 @@ declare interface BaseCreateAgentInteractionParams {
676
700
  * Body param: Enforces that the generated response is a JSON object that complies
677
701
  * with the JSON schema specified in this field.
678
702
  */
679
- response_format?: unknown;
703
+ response_format?: Array<AudioResponseFormat | TextResponseFormat | ImageResponseFormat | VideoResponseFormat | unknown> | AudioResponseFormat | TextResponseFormat | ImageResponseFormat | VideoResponseFormat | unknown;
680
704
  /**
681
705
  * Body param: The mime type of the response. This is required if response_format
682
706
  * is set.
@@ -722,7 +746,7 @@ declare interface BaseCreateModelInteractionParams {
722
746
  /**
723
747
  * Body param: The input for the interaction.
724
748
  */
725
- input: Array<Content_2> | string | Array<Turn> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | CodeExecutionCallContent | URLContextCallContent | MCPServerToolCallContent | GoogleSearchCallContent | FileSearchCallContent | GoogleMapsCallContent | FunctionResultContent | CodeExecutionResultContent | URLContextResultContent | GoogleSearchResultContent | MCPServerToolResultContent | FileSearchResultContent | GoogleMapsResultContent;
749
+ input: string | Array<Step> | Array<Content_2> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent;
726
750
  /**
727
751
  * Body param: The name of the `Model` used for generating the interaction.
728
752
  */
@@ -743,7 +767,7 @@ declare interface BaseCreateModelInteractionParams {
743
767
  * Body param: Enforces that the generated response is a JSON object that complies
744
768
  * with the JSON schema specified in this field.
745
769
  */
746
- response_format?: unknown;
770
+ response_format?: Array<AudioResponseFormat | TextResponseFormat | ImageResponseFormat | VideoResponseFormat | unknown> | AudioResponseFormat | TextResponseFormat | ImageResponseFormat | VideoResponseFormat | unknown;
747
771
  /**
748
772
  * Body param: The mime type of the response. This is required if response_format
749
773
  * is set.
@@ -1770,9 +1794,9 @@ declare interface CodeExecutionCallArguments {
1770
1794
  }
1771
1795
 
1772
1796
  /**
1773
- * Code execution content.
1797
+ * Code execution call step.
1774
1798
  */
1775
- declare interface CodeExecutionCallContent {
1799
+ declare interface CodeExecutionCallStep {
1776
1800
  /**
1777
1801
  * Required. A unique ID for this specific tool call.
1778
1802
  */
@@ -1780,7 +1804,7 @@ declare interface CodeExecutionCallContent {
1780
1804
  /**
1781
1805
  * Required. The arguments to pass to the code execution.
1782
1806
  */
1783
- arguments: CodeExecutionCallArguments;
1807
+ arguments: CodeExecutionCallStep.Arguments;
1784
1808
  type: 'code_execution_call';
1785
1809
  /**
1786
1810
  * A signature hash for backend validation.
@@ -1788,6 +1812,22 @@ declare interface CodeExecutionCallContent {
1788
1812
  signature?: string;
1789
1813
  }
1790
1814
 
1815
+ declare namespace CodeExecutionCallStep {
1816
+ /**
1817
+ * Required. The arguments to pass to the code execution.
1818
+ */
1819
+ interface Arguments {
1820
+ /**
1821
+ * The code to be executed.
1822
+ */
1823
+ code?: string;
1824
+ /**
1825
+ * Programming language of the `code`.
1826
+ */
1827
+ language?: 'python';
1828
+ }
1829
+ }
1830
+
1791
1831
  /** Result of executing the `ExecutableCode`.
1792
1832
 
1793
1833
  Generated only when the `CodeExecution` tool is used. */
@@ -1801,9 +1841,9 @@ export declare interface CodeExecutionResult {
1801
1841
  }
1802
1842
 
1803
1843
  /**
1804
- * Code execution result content.
1844
+ * Code execution result step.
1805
1845
  */
1806
- declare interface CodeExecutionResultContent {
1846
+ declare interface CodeExecutionResultStep {
1807
1847
  /**
1808
1848
  * Required. ID to match the ID from the function call block.
1809
1849
  */
@@ -1905,288 +1945,7 @@ export declare interface Content {
1905
1945
  /**
1906
1946
  * The content of the response.
1907
1947
  */
1908
- declare type Content_2 = TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | CodeExecutionCallContent | URLContextCallContent | MCPServerToolCallContent | GoogleSearchCallContent | FileSearchCallContent | GoogleMapsCallContent | FunctionResultContent | CodeExecutionResultContent | URLContextResultContent | GoogleSearchResultContent | MCPServerToolResultContent | FileSearchResultContent | GoogleMapsResultContent;
1909
-
1910
- declare interface ContentDelta {
1911
- /**
1912
- * The delta content data for a content block.
1913
- */
1914
- delta: ContentDelta.Text | ContentDelta.Image | ContentDelta.Audio | ContentDelta.Document | ContentDelta.Video | ContentDelta.ThoughtSummary | ContentDelta.ThoughtSignature | ContentDelta.FunctionCall | ContentDelta.CodeExecutionCall | ContentDelta.URLContextCall | ContentDelta.GoogleSearchCall | ContentDelta.MCPServerToolCall | ContentDelta.FileSearchCall | ContentDelta.GoogleMapsCall | ContentDelta.FunctionResult | ContentDelta.CodeExecutionResult | ContentDelta.URLContextResult | ContentDelta.GoogleSearchResult | ContentDelta.MCPServerToolResult | ContentDelta.FileSearchResult | ContentDelta.GoogleMapsResult | ContentDelta.TextAnnotation;
1915
- event_type: 'content.delta';
1916
- index: number;
1917
- /**
1918
- * The event_id token to be used to resume the interaction stream, from this event.
1919
- */
1920
- event_id?: string;
1921
- }
1922
-
1923
- declare namespace ContentDelta {
1924
- interface Text {
1925
- text: string;
1926
- type: 'text';
1927
- }
1928
- interface Image {
1929
- type: 'image';
1930
- data?: string;
1931
- mime_type?: 'image/png' | 'image/jpeg' | 'image/webp' | 'image/heic' | 'image/heif' | 'image/gif' | 'image/bmp' | 'image/tiff';
1932
- /**
1933
- * The resolution of the media.
1934
- */
1935
- resolution?: 'low' | 'medium' | 'high' | 'ultra_high';
1936
- uri?: string;
1937
- }
1938
- interface Audio {
1939
- type: 'audio';
1940
- /**
1941
- * The number of audio channels.
1942
- */
1943
- channels?: number;
1944
- data?: string;
1945
- mime_type?: 'audio/wav' | 'audio/mp3' | 'audio/aiff' | 'audio/aac' | 'audio/ogg' | 'audio/flac' | 'audio/mpeg' | 'audio/m4a' | 'audio/l16' | 'audio/opus' | 'audio/alaw' | 'audio/mulaw';
1946
- /**
1947
- * @deprecated Deprecated. Use sample_rate instead. The value is ignored.
1948
- */
1949
- rate?: number;
1950
- /**
1951
- * The sample rate of the audio.
1952
- */
1953
- sample_rate?: number;
1954
- uri?: string;
1955
- }
1956
- interface Document {
1957
- type: 'document';
1958
- data?: string;
1959
- mime_type?: 'application/pdf';
1960
- uri?: string;
1961
- }
1962
- interface Video {
1963
- type: 'video';
1964
- data?: string;
1965
- mime_type?: 'video/mp4' | 'video/mpeg' | 'video/mpg' | 'video/mov' | 'video/avi' | 'video/x-flv' | 'video/webm' | 'video/wmv' | 'video/3gpp';
1966
- /**
1967
- * The resolution of the media.
1968
- */
1969
- resolution?: 'low' | 'medium' | 'high' | 'ultra_high';
1970
- uri?: string;
1971
- }
1972
- interface ThoughtSummary {
1973
- type: 'thought_summary';
1974
- /**
1975
- * A new summary item to be added to the thought.
1976
- */
1977
- content?: InteractionsAPI.TextContent | InteractionsAPI.ImageContent;
1978
- }
1979
- interface ThoughtSignature {
1980
- type: 'thought_signature';
1981
- /**
1982
- * Signature to match the backend source to be part of the generation.
1983
- */
1984
- signature?: string;
1985
- }
1986
- interface FunctionCall {
1987
- /**
1988
- * Required. A unique ID for this specific tool call.
1989
- */
1990
- id: string;
1991
- arguments: {
1992
- [key: string]: unknown;
1993
- };
1994
- name: string;
1995
- type: 'function_call';
1996
- /**
1997
- * A signature hash for backend validation.
1998
- */
1999
- signature?: string;
2000
- }
2001
- interface CodeExecutionCall {
2002
- /**
2003
- * Required. A unique ID for this specific tool call.
2004
- */
2005
- id: string;
2006
- /**
2007
- * The arguments to pass to the code execution.
2008
- */
2009
- arguments: InteractionsAPI.CodeExecutionCallArguments;
2010
- type: 'code_execution_call';
2011
- /**
2012
- * A signature hash for backend validation.
2013
- */
2014
- signature?: string;
2015
- }
2016
- interface URLContextCall {
2017
- /**
2018
- * Required. A unique ID for this specific tool call.
2019
- */
2020
- id: string;
2021
- /**
2022
- * The arguments to pass to the URL context.
2023
- */
2024
- arguments: InteractionsAPI.URLContextCallArguments;
2025
- type: 'url_context_call';
2026
- /**
2027
- * A signature hash for backend validation.
2028
- */
2029
- signature?: string;
2030
- }
2031
- interface GoogleSearchCall {
2032
- /**
2033
- * Required. A unique ID for this specific tool call.
2034
- */
2035
- id: string;
2036
- /**
2037
- * The arguments to pass to Google Search.
2038
- */
2039
- arguments: InteractionsAPI.GoogleSearchCallArguments;
2040
- type: 'google_search_call';
2041
- /**
2042
- * A signature hash for backend validation.
2043
- */
2044
- signature?: string;
2045
- }
2046
- interface MCPServerToolCall {
2047
- /**
2048
- * Required. A unique ID for this specific tool call.
2049
- */
2050
- id: string;
2051
- arguments: {
2052
- [key: string]: unknown;
2053
- };
2054
- name: string;
2055
- server_name: string;
2056
- type: 'mcp_server_tool_call';
2057
- /**
2058
- * A signature hash for backend validation.
2059
- */
2060
- signature?: string;
2061
- }
2062
- interface FileSearchCall {
2063
- /**
2064
- * Required. A unique ID for this specific tool call.
2065
- */
2066
- id: string;
2067
- type: 'file_search_call';
2068
- /**
2069
- * A signature hash for backend validation.
2070
- */
2071
- signature?: string;
2072
- }
2073
- interface GoogleMapsCall {
2074
- /**
2075
- * Required. A unique ID for this specific tool call.
2076
- */
2077
- id: string;
2078
- type: 'google_maps_call';
2079
- /**
2080
- * The arguments to pass to the Google Maps tool.
2081
- */
2082
- arguments?: InteractionsAPI.GoogleMapsCallArguments;
2083
- /**
2084
- * A signature hash for backend validation.
2085
- */
2086
- signature?: string;
2087
- }
2088
- interface FunctionResult {
2089
- /**
2090
- * Required. ID to match the ID from the function call block.
2091
- */
2092
- call_id: string;
2093
- result: unknown | Array<InteractionsAPI.TextContent | InteractionsAPI.ImageContent> | string;
2094
- type: 'function_result';
2095
- is_error?: boolean;
2096
- name?: string;
2097
- /**
2098
- * A signature hash for backend validation.
2099
- */
2100
- signature?: string;
2101
- }
2102
- interface CodeExecutionResult {
2103
- /**
2104
- * Required. ID to match the ID from the function call block.
2105
- */
2106
- call_id: string;
2107
- result: string;
2108
- type: 'code_execution_result';
2109
- is_error?: boolean;
2110
- /**
2111
- * A signature hash for backend validation.
2112
- */
2113
- signature?: string;
2114
- }
2115
- interface URLContextResult {
2116
- /**
2117
- * Required. ID to match the ID from the function call block.
2118
- */
2119
- call_id: string;
2120
- result: Array<InteractionsAPI.URLContextResult>;
2121
- type: 'url_context_result';
2122
- is_error?: boolean;
2123
- /**
2124
- * A signature hash for backend validation.
2125
- */
2126
- signature?: string;
2127
- }
2128
- interface GoogleSearchResult {
2129
- /**
2130
- * Required. ID to match the ID from the function call block.
2131
- */
2132
- call_id: string;
2133
- result: Array<InteractionsAPI.GoogleSearchResult>;
2134
- type: 'google_search_result';
2135
- is_error?: boolean;
2136
- /**
2137
- * A signature hash for backend validation.
2138
- */
2139
- signature?: string;
2140
- }
2141
- interface MCPServerToolResult {
2142
- /**
2143
- * Required. ID to match the ID from the function call block.
2144
- */
2145
- call_id: string;
2146
- result: unknown | Array<InteractionsAPI.TextContent | InteractionsAPI.ImageContent> | string;
2147
- type: 'mcp_server_tool_result';
2148
- name?: string;
2149
- server_name?: string;
2150
- /**
2151
- * A signature hash for backend validation.
2152
- */
2153
- signature?: string;
2154
- }
2155
- interface FileSearchResult {
2156
- /**
2157
- * Required. ID to match the ID from the function call block.
2158
- */
2159
- call_id: string;
2160
- result: Array<unknown>;
2161
- type: 'file_search_result';
2162
- /**
2163
- * A signature hash for backend validation.
2164
- */
2165
- signature?: string;
2166
- }
2167
- interface GoogleMapsResult {
2168
- /**
2169
- * Required. ID to match the ID from the function call block.
2170
- */
2171
- call_id: string;
2172
- type: 'google_maps_result';
2173
- /**
2174
- * The results of the Google Maps.
2175
- */
2176
- result?: Array<InteractionsAPI.GoogleMapsResult>;
2177
- /**
2178
- * A signature hash for backend validation.
2179
- */
2180
- signature?: string;
2181
- }
2182
- interface TextAnnotation {
2183
- type: 'text_annotation';
2184
- /**
2185
- * Citation information for model-generated content.
2186
- */
2187
- annotations?: Array<InteractionsAPI.Annotation>;
2188
- }
2189
- }
1948
+ declare type Content_2 = TextContent | ImageContent | AudioContent | DocumentContent | VideoContent;
2190
1949
 
2191
1950
  /** The embedding generated from an input content. */
2192
1951
  export declare interface ContentEmbedding {
@@ -2228,28 +1987,6 @@ export declare class ContentReferenceImage {
2228
1987
  toReferenceImageAPI(): ReferenceImageAPIInternal;
2229
1988
  }
2230
1989
 
2231
- declare interface ContentStart {
2232
- /**
2233
- * The content of the response.
2234
- */
2235
- content: Content_2;
2236
- event_type: 'content.start';
2237
- index: number;
2238
- /**
2239
- * The event_id token to be used to resume the interaction stream, from this event.
2240
- */
2241
- event_id?: string;
2242
- }
2243
-
2244
- declare interface ContentStop {
2245
- event_type: 'content.stop';
2246
- index: number;
2247
- /**
2248
- * The event_id token to be used to resume the interaction stream, from this event.
2249
- */
2250
- event_id?: string;
2251
- }
2252
-
2253
1990
  export declare type ContentUnion = Content | PartUnion[] | PartUnion;
2254
1991
 
2255
1992
  /** Enables context window compression -- mechanism managing model context window so it does not exceed given length. */
@@ -2595,6 +2332,10 @@ export declare interface CreateFileSearchStoreConfig {
2595
2332
  /** The human-readable display name for the file search store.
2596
2333
  */
2597
2334
  displayName?: string;
2335
+ /** The embedding model to use for the FileSearchStore.
2336
+ Format: `models/{model}`. If not specified, the default embedding model will be used.
2337
+ */
2338
+ embeddingModel?: string;
2598
2339
  }
2599
2340
 
2600
2341
  /** Config for file_search_stores.create parameters. */
@@ -3211,6 +2952,19 @@ export declare interface DownloadFileParameters {
3211
2952
  config?: DownloadFileConfig;
3212
2953
  }
3213
2954
 
2955
+ /** Used to override the default configuration. */
2956
+ export declare interface DownloadMediaConfig {
2957
+ /** Used to override HTTP request options. */
2958
+ httpOptions?: HttpOptions;
2959
+ /** Abort signal which can be used to cancel the request.
2960
+
2961
+ NOTE: AbortSignal is a client-only operation. Using it to cancel an
2962
+ operation will not cancel the request in the service. You will still
2963
+ be charged usage for any applicable operations.
2964
+ */
2965
+ abortSignal?: AbortSignal;
2966
+ }
2967
+
3214
2968
  /**
3215
2969
  * Configuration for dynamic agents.
3216
2970
  */
@@ -3893,9 +3647,9 @@ export declare interface FileSearch {
3893
3647
  }
3894
3648
 
3895
3649
  /**
3896
- * File Search content.
3650
+ * File Search call step.
3897
3651
  */
3898
- declare interface FileSearchCallContent {
3652
+ declare interface FileSearchCallStep {
3899
3653
  /**
3900
3654
  * Required. A unique ID for this specific tool call.
3901
3655
  */
@@ -3908,18 +3662,14 @@ declare interface FileSearchCallContent {
3908
3662
  }
3909
3663
 
3910
3664
  /**
3911
- * File Search result content.
3665
+ * File Search result step.
3912
3666
  */
3913
- declare interface FileSearchResultContent {
3667
+ declare interface FileSearchResultStep {
3914
3668
  /**
3915
3669
  * Required. ID to match the ID from the function call block.
3916
3670
  */
3917
3671
  call_id: string;
3918
3672
  type: 'file_search_result';
3919
- /**
3920
- * Optional. The results of the File Search.
3921
- */
3922
- result?: Array<unknown>;
3923
3673
  /**
3924
3674
  * A signature hash for backend validation.
3925
3675
  */
@@ -3945,6 +3695,8 @@ export declare interface FileSearchStore {
3945
3695
  /** The size of raw bytes ingested into the FileSearchStore. This is the
3946
3696
  total size of all the documents in the FileSearchStore. */
3947
3697
  sizeBytes?: string;
3698
+ /** The embedding model used by the FileSearchStore. */
3699
+ embeddingModel?: string;
3948
3700
  }
3949
3701
 
3950
3702
  declare class FileSearchStores extends BaseModule {
@@ -4003,6 +3755,15 @@ declare class FileSearchStores extends BaseModule {
4003
3755
  * ```
4004
3756
  */
4005
3757
  uploadToFileSearchStore(params: types.UploadToFileSearchStoreParameters): Promise<types.UploadToFileSearchStoreOperation>;
3758
+ /**
3759
+ * Downloads media using a Media ID or URI.
3760
+ * This method is only supported in the Gemini Developer client.
3761
+ *
3762
+ * @param uri - The URI or Media ID of the blob.
3763
+ * @param config - Optional configuration for the download.
3764
+ * @returns A promise that resolves to the blob data as a Uint8Array.
3765
+ */
3766
+ downloadMedia(uri: string, config?: types.DownloadMediaConfig): Promise<Uint8Array>;
4006
3767
  /**
4007
3768
  * Creates a File Search Store.
4008
3769
  *
@@ -4206,31 +3967,6 @@ export declare interface FunctionCall {
4206
3967
  willContinue?: boolean;
4207
3968
  }
4208
3969
 
4209
- /**
4210
- * A function tool call content block.
4211
- */
4212
- declare interface FunctionCallContent {
4213
- /**
4214
- * Required. A unique ID for this specific tool call.
4215
- */
4216
- id: string;
4217
- /**
4218
- * Required. The arguments to pass to the function.
4219
- */
4220
- arguments: {
4221
- [key: string]: unknown;
4222
- };
4223
- /**
4224
- * Required. The name of the tool to call.
4225
- */
4226
- name: string;
4227
- type: 'function_call';
4228
- /**
4229
- * A signature hash for backend validation.
4230
- */
4231
- signature?: string;
4232
- }
4233
-
4234
3970
  /** Function calling config. */
4235
3971
  export declare interface FunctionCallingConfig {
4236
3972
  /** Optional. Function names to call. Only set when the Mode is ANY. Function names should match [FunctionDeclaration.name]. With mode set to ANY, model will predict a function call from the set of function names provided. */
@@ -4265,6 +4001,31 @@ export declare enum FunctionCallingConfigMode {
4265
4001
  VALIDATED = "VALIDATED"
4266
4002
  }
4267
4003
 
4004
+ /**
4005
+ * A function tool call step.
4006
+ */
4007
+ declare interface FunctionCallStep {
4008
+ /**
4009
+ * Required. A unique ID for this specific tool call.
4010
+ */
4011
+ id: string;
4012
+ /**
4013
+ * Required. The arguments to pass to the function.
4014
+ */
4015
+ arguments: {
4016
+ [key: string]: unknown;
4017
+ };
4018
+ /**
4019
+ * Required. The name of the tool to call.
4020
+ */
4021
+ name: string;
4022
+ type: 'function_call';
4023
+ /**
4024
+ * A signature hash for backend validation.
4025
+ */
4026
+ signature?: string;
4027
+ }
4028
+
4268
4029
  /** Structured representation of a function declaration as defined by the [OpenAPI 3.0 specification](https://spec.openapis.org/oas/v3.0.3). Included in this declaration are the function name, description, parameters and response type. This FunctionDeclaration is a representation of a block of code that can be used as a `Tool` by the model and executed by the client. */
4269
4030
  export declare interface FunctionDeclaration {
4270
4031
  /** Optional. Description and purpose of the function. Model uses it to decide how and whether to call the function. */
@@ -4363,9 +4124,9 @@ export declare enum FunctionResponseScheduling {
4363
4124
  }
4364
4125
 
4365
4126
  /**
4366
- * A function tool result content block.
4127
+ * Result of a function tool call.
4367
4128
  */
4368
- declare interface FunctionResultContent {
4129
+ declare interface FunctionResultStep {
4369
4130
  /**
4370
4131
  * Required. ID to match the ID from the function call block.
4371
4132
  */
@@ -5492,9 +5253,9 @@ declare interface GoogleMapsCallArguments {
5492
5253
  }
5493
5254
 
5494
5255
  /**
5495
- * Google Maps content.
5256
+ * Google Maps call step.
5496
5257
  */
5497
- declare interface GoogleMapsCallContent {
5258
+ declare interface GoogleMapsCallStep {
5498
5259
  /**
5499
5260
  * Required. A unique ID for this specific tool call.
5500
5261
  */
@@ -5503,13 +5264,25 @@ declare interface GoogleMapsCallContent {
5503
5264
  /**
5504
5265
  * The arguments to pass to the Google Maps tool.
5505
5266
  */
5506
- arguments?: GoogleMapsCallArguments;
5267
+ arguments?: GoogleMapsCallStep.Arguments;
5507
5268
  /**
5508
5269
  * A signature hash for backend validation.
5509
5270
  */
5510
5271
  signature?: string;
5511
5272
  }
5512
5273
 
5274
+ declare namespace GoogleMapsCallStep {
5275
+ /**
5276
+ * The arguments to pass to the Google Maps tool.
5277
+ */
5278
+ interface Arguments {
5279
+ /**
5280
+ * The queries to be executed.
5281
+ */
5282
+ queries?: Array<string>;
5283
+ }
5284
+ }
5285
+
5513
5286
  /**
5514
5287
  * The result of the Google Maps.
5515
5288
  */
@@ -5567,17 +5340,14 @@ declare namespace GoogleMapsResult {
5567
5340
  }
5568
5341
 
5569
5342
  /**
5570
- * Google Maps result content.
5343
+ * Google Maps result step.
5571
5344
  */
5572
- declare interface GoogleMapsResultContent {
5345
+ declare interface GoogleMapsResultStep {
5573
5346
  /**
5574
5347
  * Required. ID to match the ID from the function call block.
5575
5348
  */
5576
5349
  call_id: string;
5577
- /**
5578
- * Required. The results of the Google Maps.
5579
- */
5580
- result: Array<GoogleMapsResult>;
5350
+ result: Array<GoogleMapsResultStep.Result>;
5581
5351
  type: 'google_maps_result';
5582
5352
  /**
5583
5353
  * A signature hash for backend validation.
@@ -5585,6 +5355,44 @@ declare interface GoogleMapsResultContent {
5585
5355
  signature?: string;
5586
5356
  }
5587
5357
 
5358
+ declare namespace GoogleMapsResultStep {
5359
+ /**
5360
+ * The result of the Google Maps.
5361
+ */
5362
+ interface Result {
5363
+ places?: Array<Result.Place>;
5364
+ widget_context_token?: string;
5365
+ }
5366
+ namespace Result {
5367
+ interface Place {
5368
+ name?: string;
5369
+ place_id?: string;
5370
+ review_snippets?: Array<Place.ReviewSnippet>;
5371
+ url?: string;
5372
+ }
5373
+ namespace Place {
5374
+ /**
5375
+ * Encapsulates a snippet of a user review that answers a question about the
5376
+ * features of a specific place in Google Maps.
5377
+ */
5378
+ interface ReviewSnippet {
5379
+ /**
5380
+ * The ID of the review snippet.
5381
+ */
5382
+ review_id?: string;
5383
+ /**
5384
+ * Title of the review.
5385
+ */
5386
+ title?: string;
5387
+ /**
5388
+ * A link that corresponds to the user review on Google Maps.
5389
+ */
5390
+ url?: string;
5391
+ }
5392
+ }
5393
+ }
5394
+ }
5395
+
5588
5396
  /** The `Status` type defines a logical error model that is suitable for different programming environments, including REST APIs and RPC APIs. It is used by [gRPC](https://github.com/grpc). Each `Status` message contains three pieces of data: error code, error message, and error details. You can find out more about this error model and how to work with it in the [API Design Guide](https://cloud.google.com/apis/design/errors). This data type is not supported in Gemini API. */
5589
5397
  export declare interface GoogleRpcStatus {
5590
5398
  /** The status code, which should be an enum value of google.rpc.Code. */
@@ -5618,9 +5426,9 @@ declare interface GoogleSearchCallArguments {
5618
5426
  }
5619
5427
 
5620
5428
  /**
5621
- * Google Search content.
5429
+ * Google Search call step.
5622
5430
  */
5623
- declare interface GoogleSearchCallContent {
5431
+ declare interface GoogleSearchCallStep {
5624
5432
  /**
5625
5433
  * Required. A unique ID for this specific tool call.
5626
5434
  */
@@ -5628,7 +5436,7 @@ declare interface GoogleSearchCallContent {
5628
5436
  /**
5629
5437
  * Required. The arguments to pass to Google Search.
5630
5438
  */
5631
- arguments: GoogleSearchCallArguments;
5439
+ arguments: GoogleSearchCallStep.Arguments;
5632
5440
  type: 'google_search_call';
5633
5441
  /**
5634
5442
  * The type of search grounding enabled.
@@ -5640,6 +5448,18 @@ declare interface GoogleSearchCallContent {
5640
5448
  signature?: string;
5641
5449
  }
5642
5450
 
5451
+ declare namespace GoogleSearchCallStep {
5452
+ /**
5453
+ * Required. The arguments to pass to Google Search.
5454
+ */
5455
+ interface Arguments {
5456
+ /**
5457
+ * Web search queries for the following-up web search.
5458
+ */
5459
+ queries?: Array<string>;
5460
+ }
5461
+ }
5462
+
5643
5463
  /**
5644
5464
  * The result of the Google Search.
5645
5465
  */
@@ -5651,9 +5471,9 @@ declare interface GoogleSearchResult {
5651
5471
  }
5652
5472
 
5653
5473
  /**
5654
- * Google Search result content.
5474
+ * Google Search result step.
5655
5475
  */
5656
- declare interface GoogleSearchResultContent {
5476
+ declare interface GoogleSearchResultStep {
5657
5477
  /**
5658
5478
  * Required. ID to match the ID from the function call block.
5659
5479
  */
@@ -5661,7 +5481,7 @@ declare interface GoogleSearchResultContent {
5661
5481
  /**
5662
5482
  * Required. The results of the Google Search.
5663
5483
  */
5664
- result: Array<GoogleSearchResult>;
5484
+ result: Array<GoogleSearchResultStep.Result>;
5665
5485
  type: 'google_search_result';
5666
5486
  /**
5667
5487
  * Whether the Google Search resulted in an error.
@@ -5673,6 +5493,18 @@ declare interface GoogleSearchResultContent {
5673
5493
  signature?: string;
5674
5494
  }
5675
5495
 
5496
+ declare namespace GoogleSearchResultStep {
5497
+ /**
5498
+ * The result of the Google Search.
5499
+ */
5500
+ interface Result {
5501
+ /**
5502
+ * Web content snippet that can be embedded in a web page or an app webview.
5503
+ */
5504
+ search_suggestions?: string;
5505
+ }
5506
+ }
5507
+
5676
5508
  /** Tool to retrieve public web data for grounding, powered by Google. */
5677
5509
  export declare interface GoogleSearchRetrieval {
5678
5510
  /** Specifies the dynamic retrieval configuration for the given source. */
@@ -5826,6 +5658,8 @@ export declare interface GroundingChunkRetrievedContext {
5826
5658
  fileSearchStore?: string;
5827
5659
  /** Optional. Page number of the retrieved context. This field is not supported in Vertex AI. */
5828
5660
  pageNumber?: number;
5661
+ /** Optional. Media ID. This field is not supported in Vertex AI. */
5662
+ mediaId?: string;
5829
5663
  }
5830
5664
 
5831
5665
  /** A list of string values. This data type is not supported in Vertex AI. */
@@ -6288,6 +6122,29 @@ export declare enum ImageResizeMode {
6288
6122
  PAD = "PAD"
6289
6123
  }
6290
6124
 
6125
+ /**
6126
+ * Configuration for image output format.
6127
+ */
6128
+ declare interface ImageResponseFormat {
6129
+ type: 'image';
6130
+ /**
6131
+ * The aspect ratio for the image output.
6132
+ */
6133
+ aspectRatio?: '1:1' | '2:3' | '3:2' | '3:4' | '4:3' | '4:5' | '5:4' | '9:16' | '16:9' | '21:9' | '1:8' | '8:1' | '1:4' | '4:1';
6134
+ /**
6135
+ * The delivery mode for the image output.
6136
+ */
6137
+ delivery?: 'inline' | 'url';
6138
+ /**
6139
+ * The size of the image output.
6140
+ */
6141
+ imageSize?: '512' | '1K' | '2K' | '4K';
6142
+ /**
6143
+ * The MIME type of the image output.
6144
+ */
6145
+ mimeType?: 'image/jpeg';
6146
+ }
6147
+
6291
6148
  /** Image search for grounding and related configurations. */
6292
6149
  export declare interface ImageSearch {
6293
6150
  }
@@ -6419,15 +6276,11 @@ declare interface Interaction {
6419
6276
  /**
6420
6277
  * The input for the interaction.
6421
6278
  */
6422
- input?: Array<Content_2> | string | Array<Turn> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent | ThoughtContent | FunctionCallContent | CodeExecutionCallContent | URLContextCallContent | MCPServerToolCallContent | GoogleSearchCallContent | FileSearchCallContent | GoogleMapsCallContent | FunctionResultContent | CodeExecutionResultContent | URLContextResultContent | GoogleSearchResultContent | MCPServerToolResultContent | FileSearchResultContent | GoogleMapsResultContent;
6279
+ input?: string | Array<Step> | Array<Content_2> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent;
6423
6280
  /**
6424
6281
  * The name of the `Model` used for generating the interaction.
6425
6282
  */
6426
6283
  model?: Model_2;
6427
- /**
6428
- * Output only. Responses from the model.
6429
- */
6430
- outputs?: Array<Content_2>;
6431
6284
  /**
6432
6285
  * The ID of the previous interaction, if any.
6433
6286
  */
@@ -6436,7 +6289,7 @@ declare interface Interaction {
6436
6289
  * Enforces that the generated response is a JSON object that complies with the
6437
6290
  * JSON schema specified in this field.
6438
6291
  */
6439
- response_format?: unknown;
6292
+ response_format?: Array<AudioResponseFormat | TextResponseFormat | ImageResponseFormat | VideoResponseFormat | unknown> | AudioResponseFormat | TextResponseFormat | ImageResponseFormat | VideoResponseFormat | unknown;
6440
6293
  /**
6441
6294
  * The mime type of the response. This is required if response_format is set.
6442
6295
  */
@@ -6446,13 +6299,17 @@ declare interface Interaction {
6446
6299
  */
6447
6300
  response_modalities?: Array<'text' | 'image' | 'audio' | 'video' | 'document'>;
6448
6301
  /**
6449
- * Output only. The role of the interaction.
6302
+ * @deprecated Output only. The role of the interaction.
6450
6303
  */
6451
6304
  role?: string;
6452
6305
  /**
6453
6306
  * The service tier for the interaction.
6454
6307
  */
6455
6308
  service_tier?: 'flex' | 'standard' | 'priority';
6309
+ /**
6310
+ * Output only. The steps that make up the interaction.
6311
+ */
6312
+ steps?: Array<Step>;
6456
6313
  /**
6457
6314
  * System instruction for the interaction.
6458
6315
  */
@@ -6479,8 +6336,8 @@ declare interface InteractionCancelParams {
6479
6336
  api_version?: string;
6480
6337
  }
6481
6338
 
6482
- declare interface InteractionCompleteEvent {
6483
- event_type: 'interaction.complete';
6339
+ declare interface InteractionCompletedEvent {
6340
+ event_type: 'interaction.completed';
6484
6341
  /**
6485
6342
  * Required. The completed interaction with empty outputs to reduce the payload
6486
6343
  * size. Use the preceding ContentDelta events for the actual output.
@@ -6492,6 +6349,18 @@ declare interface InteractionCompleteEvent {
6492
6349
  event_id?: string;
6493
6350
  }
6494
6351
 
6352
+ declare interface InteractionCreatedEvent {
6353
+ event_type: 'interaction.created';
6354
+ /**
6355
+ * The Interaction resource.
6356
+ */
6357
+ interaction: Interaction;
6358
+ /**
6359
+ * The event_id token to be used to resume the interaction stream, from this event.
6360
+ */
6361
+ event_id?: string;
6362
+ }
6363
+
6495
6364
  declare type InteractionCreateParams = CreateModelInteractionParamsNonStreaming | CreateModelInteractionParamsStreaming | CreateAgentInteractionParamsNonStreaming | CreateAgentInteractionParamsStreaming;
6496
6365
 
6497
6366
  declare interface InteractionDeleteParams {
@@ -6552,7 +6421,7 @@ export declare class Interactions extends BaseInteractions {
6552
6421
  }
6553
6422
 
6554
6423
  export declare namespace Interactions {
6555
- export { type AllowedTools as AllowedTools, type Annotation as Annotation, type AudioContent as AudioContent, type CodeExecutionCallArguments as CodeExecutionCallArguments, type CodeExecutionCallContent as CodeExecutionCallContent, type CodeExecutionResultContent as CodeExecutionResultContent, type Content_2 as Content, type ContentDelta as ContentDelta, type ContentStart as ContentStart, type ContentStop as ContentStop, type DeepResearchAgentConfig as DeepResearchAgentConfig, type DocumentContent as DocumentContent, type DynamicAgentConfig as DynamicAgentConfig, type ErrorEvent_2 as ErrorEvent, type FileCitation as FileCitation, type FileSearchCallContent as FileSearchCallContent, type FileSearchResultContent as FileSearchResultContent, type Function_2 as Function, type FunctionCallContent as FunctionCallContent, type FunctionResultContent as FunctionResultContent, type GenerationConfig_2 as GenerationConfig, type GoogleMapsCallArguments as GoogleMapsCallArguments, type GoogleMapsCallContent as GoogleMapsCallContent, type GoogleMapsResult as GoogleMapsResult, type GoogleMapsResultContent as GoogleMapsResultContent, type GoogleSearchCallArguments as GoogleSearchCallArguments, type GoogleSearchCallContent as GoogleSearchCallContent, type GoogleSearchResult as GoogleSearchResult, type GoogleSearchResultContent as GoogleSearchResultContent, type ImageConfig_2 as ImageConfig, type ImageContent as ImageContent, type Interaction as Interaction, type InteractionCompleteEvent as InteractionCompleteEvent, type InteractionSSEEvent as InteractionSSEEvent, type InteractionStartEvent as InteractionStartEvent, type InteractionStatusUpdate as InteractionStatusUpdate, type MCPServerToolCallContent as MCPServerToolCallContent, type MCPServerToolResultContent as MCPServerToolResultContent, type Model_2 as Model, type PlaceCitation as PlaceCitation, type SpeechConfig_2 as SpeechConfig, type TextContent as TextContent, type ThinkingLevel_2 as ThinkingLevel, type ThoughtContent as ThoughtContent, type Tool_2 as Tool, type ToolChoiceConfig as ToolChoiceConfig, type ToolChoiceType as ToolChoiceType, type Turn as Turn, type URLCitation as URLCitation, type URLContextCallArguments as URLContextCallArguments, type URLContextCallContent as URLContextCallContent, type URLContextResult as URLContextResult, type URLContextResultContent as URLContextResultContent, type Usage as Usage, type VideoContent as VideoContent, type WebhookConfig_2 as WebhookConfig, type InteractionDeleteResponse as InteractionDeleteResponse, type InteractionCreateParams as InteractionCreateParams, type CreateModelInteractionParamsNonStreaming as CreateModelInteractionParamsNonStreaming, type CreateModelInteractionParamsStreaming as CreateModelInteractionParamsStreaming, type CreateAgentInteractionParamsNonStreaming as CreateAgentInteractionParamsNonStreaming, type CreateAgentInteractionParamsStreaming as CreateAgentInteractionParamsStreaming, type InteractionDeleteParams as InteractionDeleteParams, type InteractionCancelParams as InteractionCancelParams, type InteractionGetParams as InteractionGetParams, type InteractionGetParamsNonStreaming as InteractionGetParamsNonStreaming, type InteractionGetParamsStreaming as InteractionGetParamsStreaming, };
6424
+ export { type AllowedTools as AllowedTools, type Annotation as Annotation, type AudioContent as AudioContent, type AudioResponseFormat as AudioResponseFormat, type CodeExecutionCallArguments as CodeExecutionCallArguments, type CodeExecutionCallStep as CodeExecutionCallStep, type CodeExecutionResultStep as CodeExecutionResultStep, type Content_2 as Content, type DeepResearchAgentConfig as DeepResearchAgentConfig, type DocumentContent as DocumentContent, type DynamicAgentConfig as DynamicAgentConfig, type ErrorEvent_2 as ErrorEvent, type FileCitation as FileCitation, type FileSearchCallStep as FileSearchCallStep, type FileSearchResultStep as FileSearchResultStep, type Function_2 as Function, type FunctionCallStep as FunctionCallStep, type FunctionResultStep as FunctionResultStep, type GenerationConfig_2 as GenerationConfig, type GoogleMapsCallArguments as GoogleMapsCallArguments, type GoogleMapsCallStep as GoogleMapsCallStep, type GoogleMapsResult as GoogleMapsResult, type GoogleMapsResultStep as GoogleMapsResultStep, type GoogleSearchCallArguments as GoogleSearchCallArguments, type GoogleSearchCallStep as GoogleSearchCallStep, type GoogleSearchResult as GoogleSearchResult, type GoogleSearchResultStep as GoogleSearchResultStep, type ImageConfig_2 as ImageConfig, type ImageContent as ImageContent, type ImageResponseFormat as ImageResponseFormat, type Interaction as Interaction, type InteractionCompletedEvent as InteractionCompletedEvent, type InteractionCreatedEvent as InteractionCreatedEvent, type InteractionSSEEvent as InteractionSSEEvent, type InteractionStatusUpdate as InteractionStatusUpdate, type MCPServerToolCallStep as MCPServerToolCallStep, type MCPServerToolResultStep as MCPServerToolResultStep, type Model_2 as Model, type ModelOutputStep as ModelOutputStep, type PlaceCitation as PlaceCitation, type SpeechConfig_2 as SpeechConfig, type Step as Step, type StepDelta as StepDelta, type StepStart as StepStart, type StepStop as StepStop, type TextContent as TextContent, type TextResponseFormat as TextResponseFormat, type ThinkingLevel_2 as ThinkingLevel, type ThoughtStep as ThoughtStep, type Tool_2 as Tool, type ToolChoiceConfig as ToolChoiceConfig, type ToolChoiceType as ToolChoiceType, type URLCitation as URLCitation, type URLContextCallArguments as URLContextCallArguments, type URLContextCallStep as URLContextCallStep, type URLContextResult as URLContextResult, type URLContextResultStep as URLContextResultStep, type Usage as Usage, type UserInputStep as UserInputStep, type VideoContent as VideoContent, type VideoResponseFormat as VideoResponseFormat, type WebhookConfig_2 as WebhookConfig, type InteractionDeleteResponse as InteractionDeleteResponse, type InteractionCreateParams as InteractionCreateParams, type CreateModelInteractionParamsNonStreaming as CreateModelInteractionParamsNonStreaming, type CreateModelInteractionParamsStreaming as CreateModelInteractionParamsStreaming, type CreateAgentInteractionParamsNonStreaming as CreateAgentInteractionParamsNonStreaming, type CreateAgentInteractionParamsStreaming as CreateAgentInteractionParamsStreaming, type InteractionDeleteParams as InteractionDeleteParams, type InteractionCancelParams as InteractionCancelParams, type InteractionGetParams as InteractionGetParams, type InteractionGetParamsNonStreaming as InteractionGetParamsNonStreaming, type InteractionGetParamsStreaming as InteractionGetParamsStreaming, };
6556
6425
  }
6557
6426
 
6558
6427
  declare namespace InteractionsAPI {
@@ -6562,58 +6431,64 @@ declare namespace InteractionsAPI {
6562
6431
  AllowedTools,
6563
6432
  Annotation,
6564
6433
  AudioContent,
6434
+ AudioResponseFormat,
6565
6435
  CodeExecutionCallArguments,
6566
- CodeExecutionCallContent,
6567
- CodeExecutionResultContent,
6436
+ CodeExecutionCallStep,
6437
+ CodeExecutionResultStep,
6568
6438
  Content_2 as Content,
6569
- ContentDelta,
6570
- ContentStart,
6571
- ContentStop,
6572
6439
  DeepResearchAgentConfig,
6573
6440
  DocumentContent,
6574
6441
  DynamicAgentConfig,
6575
6442
  ErrorEvent_2 as ErrorEvent,
6576
6443
  FileCitation,
6577
- FileSearchCallContent,
6578
- FileSearchResultContent,
6444
+ FileSearchCallStep,
6445
+ FileSearchResultStep,
6579
6446
  Function_2 as Function,
6580
- FunctionCallContent,
6581
- FunctionResultContent,
6447
+ FunctionCallStep,
6448
+ FunctionResultStep,
6582
6449
  GenerationConfig_2 as GenerationConfig,
6583
6450
  GoogleMapsCallArguments,
6584
- GoogleMapsCallContent,
6451
+ GoogleMapsCallStep,
6585
6452
  GoogleMapsResult,
6586
- GoogleMapsResultContent,
6453
+ GoogleMapsResultStep,
6587
6454
  GoogleSearchCallArguments,
6588
- GoogleSearchCallContent,
6455
+ GoogleSearchCallStep,
6589
6456
  GoogleSearchResult,
6590
- GoogleSearchResultContent,
6457
+ GoogleSearchResultStep,
6591
6458
  ImageConfig_2 as ImageConfig,
6592
6459
  ImageContent,
6460
+ ImageResponseFormat,
6593
6461
  Interaction,
6594
- InteractionCompleteEvent,
6462
+ InteractionCompletedEvent,
6463
+ InteractionCreatedEvent,
6595
6464
  InteractionSSEEvent,
6596
- InteractionStartEvent,
6597
6465
  InteractionStatusUpdate,
6598
- MCPServerToolCallContent,
6599
- MCPServerToolResultContent,
6466
+ MCPServerToolCallStep,
6467
+ MCPServerToolResultStep,
6600
6468
  Model_2 as Model,
6469
+ ModelOutputStep,
6601
6470
  PlaceCitation,
6602
6471
  SpeechConfig_2 as SpeechConfig,
6472
+ Step,
6473
+ StepDelta,
6474
+ StepStart,
6475
+ StepStop,
6603
6476
  TextContent,
6477
+ TextResponseFormat,
6604
6478
  ThinkingLevel_2 as ThinkingLevel,
6605
- ThoughtContent,
6479
+ ThoughtStep,
6606
6480
  Tool_2 as Tool,
6607
6481
  ToolChoiceConfig,
6608
6482
  ToolChoiceType,
6609
- Turn,
6610
6483
  URLCitation,
6611
6484
  URLContextCallArguments,
6612
- URLContextCallContent,
6485
+ URLContextCallStep,
6613
6486
  URLContextResult,
6614
- URLContextResultContent,
6487
+ URLContextResultStep,
6615
6488
  Usage,
6489
+ UserInputStep,
6616
6490
  VideoContent,
6491
+ VideoResponseFormat,
6617
6492
  WebhookConfig_2 as WebhookConfig,
6618
6493
  InteractionDeleteResponse,
6619
6494
  InteractionCreateParams,
@@ -6632,19 +6507,7 @@ declare namespace InteractionsAPI {
6632
6507
  }
6633
6508
  }
6634
6509
 
6635
- declare type InteractionSSEEvent = InteractionStartEvent | InteractionCompleteEvent | InteractionStatusUpdate | ContentStart | ContentDelta | ContentStop | ErrorEvent_2;
6636
-
6637
- declare interface InteractionStartEvent {
6638
- event_type: 'interaction.start';
6639
- /**
6640
- * The Interaction resource.
6641
- */
6642
- interaction: Interaction;
6643
- /**
6644
- * The event_id token to be used to resume the interaction stream, from this event.
6645
- */
6646
- event_id?: string;
6647
- }
6510
+ declare type InteractionSSEEvent = InteractionCreatedEvent | InteractionCompletedEvent | InteractionStatusUpdate | ErrorEvent_2 | StepStart | StepDelta | StepStop;
6648
6511
 
6649
6512
  declare interface InteractionStatusUpdate {
6650
6513
  event_type: 'interaction.status_update';
@@ -7847,9 +7710,9 @@ export declare interface McpServer {
7847
7710
  }
7848
7711
 
7849
7712
  /**
7850
- * MCPServer tool call content.
7713
+ * MCPServer tool call step.
7851
7714
  */
7852
- declare interface MCPServerToolCallContent {
7715
+ declare interface MCPServerToolCallStep {
7853
7716
  /**
7854
7717
  * Required. A unique ID for this specific tool call.
7855
7718
  */
@@ -7876,9 +7739,9 @@ declare interface MCPServerToolCallContent {
7876
7739
  }
7877
7740
 
7878
7741
  /**
7879
- * MCPServer tool result content.
7742
+ * MCPServer tool result step.
7880
7743
  */
7881
- declare interface MCPServerToolResultContent {
7744
+ declare interface MCPServerToolResultStep {
7882
7745
  /**
7883
7746
  * Required. ID to match the ID from the function call block.
7884
7747
  */
@@ -7886,7 +7749,7 @@ declare interface MCPServerToolResultContent {
7886
7749
  /**
7887
7750
  * The output from the MCP server call. Can be simple text or rich content.
7888
7751
  */
7889
- result: unknown | Array<TextContent | ImageContent> | string;
7752
+ result: unknown | string | Array<TextContent | ImageContent>;
7890
7753
  type: 'mcp_server_tool_result';
7891
7754
  /**
7892
7755
  * Name of the tool which is called for this specific tool call.
@@ -8071,6 +7934,14 @@ export declare interface ModelArmorConfig {
8071
7934
  responseTemplateName?: string;
8072
7935
  }
8073
7936
 
7937
+ /**
7938
+ * Output generated by the model.
7939
+ */
7940
+ declare interface ModelOutputStep {
7941
+ type: 'model_output';
7942
+ content?: Array<Content_2>;
7943
+ }
7944
+
8074
7945
  export declare class Models extends BaseModule {
8075
7946
  private readonly apiClient;
8076
7947
  constructor(apiClient: ApiClient);
@@ -10098,6 +9969,119 @@ export declare enum StartSensitivity {
10098
9969
  START_SENSITIVITY_LOW = "START_SENSITIVITY_LOW"
10099
9970
  }
10100
9971
 
9972
+ /**
9973
+ * A step in the interaction.
9974
+ */
9975
+ declare type Step = UserInputStep | ModelOutputStep | ThoughtStep | FunctionCallStep | CodeExecutionCallStep | URLContextCallStep | MCPServerToolCallStep | GoogleSearchCallStep | FileSearchCallStep | GoogleMapsCallStep | FunctionResultStep | CodeExecutionResultStep | URLContextResultStep | GoogleSearchResultStep | MCPServerToolResultStep | FileSearchResultStep | GoogleMapsResultStep;
9976
+
9977
+ declare interface StepDelta {
9978
+ delta: StepDelta.Text | StepDelta.Image | StepDelta.Audio | StepDelta.Document | StepDelta.Video | StepDelta.ThoughtSummary | StepDelta.ThoughtSignature | StepDelta.TextAnnotationDelta | StepDelta.ArgumentsDelta;
9979
+ event_type: 'step.delta';
9980
+ index: number;
9981
+ /**
9982
+ * The event_id token to be used to resume the interaction stream, from this event.
9983
+ */
9984
+ event_id?: string;
9985
+ }
9986
+
9987
+ declare namespace StepDelta {
9988
+ interface Text {
9989
+ text: string;
9990
+ type: 'text';
9991
+ }
9992
+ interface Image {
9993
+ type: 'image';
9994
+ data?: string;
9995
+ mime_type?: 'image/png' | 'image/jpeg' | 'image/webp' | 'image/heic' | 'image/heif' | 'image/gif' | 'image/bmp' | 'image/tiff';
9996
+ /**
9997
+ * The resolution of the media.
9998
+ */
9999
+ resolution?: 'low' | 'medium' | 'high' | 'ultra_high';
10000
+ uri?: string;
10001
+ }
10002
+ interface Audio {
10003
+ type: 'audio';
10004
+ /**
10005
+ * The number of audio channels.
10006
+ */
10007
+ channels?: number;
10008
+ data?: string;
10009
+ mime_type?: 'audio/wav' | 'audio/mp3' | 'audio/aiff' | 'audio/aac' | 'audio/ogg' | 'audio/flac' | 'audio/mpeg' | 'audio/m4a' | 'audio/l16' | 'audio/opus' | 'audio/alaw' | 'audio/mulaw';
10010
+ /**
10011
+ * @deprecated Deprecated. Use sample_rate instead. The value is ignored.
10012
+ */
10013
+ rate?: number;
10014
+ /**
10015
+ * The sample rate of the audio.
10016
+ */
10017
+ sample_rate?: number;
10018
+ uri?: string;
10019
+ }
10020
+ interface Document {
10021
+ type: 'document';
10022
+ data?: string;
10023
+ mime_type?: 'application/pdf';
10024
+ uri?: string;
10025
+ }
10026
+ interface Video {
10027
+ type: 'video';
10028
+ data?: string;
10029
+ mime_type?: 'video/mp4' | 'video/mpeg' | 'video/mpg' | 'video/mov' | 'video/avi' | 'video/x-flv' | 'video/webm' | 'video/wmv' | 'video/3gpp';
10030
+ /**
10031
+ * The resolution of the media.
10032
+ */
10033
+ resolution?: 'low' | 'medium' | 'high' | 'ultra_high';
10034
+ uri?: string;
10035
+ }
10036
+ interface ThoughtSummary {
10037
+ type: 'thought_summary';
10038
+ /**
10039
+ * A new summary item to be added to the thought.
10040
+ */
10041
+ content?: InteractionsAPI.TextContent | InteractionsAPI.ImageContent;
10042
+ }
10043
+ interface ThoughtSignature {
10044
+ type: 'thought_signature';
10045
+ /**
10046
+ * Signature to match the backend source to be part of the generation.
10047
+ */
10048
+ signature?: string;
10049
+ }
10050
+ interface TextAnnotationDelta {
10051
+ type: 'text_annotation_delta';
10052
+ /**
10053
+ * Citation information for model-generated content.
10054
+ */
10055
+ annotations?: Array<InteractionsAPI.Annotation>;
10056
+ }
10057
+ interface ArgumentsDelta {
10058
+ type: 'arguments_delta';
10059
+ partial_arguments?: string;
10060
+ }
10061
+ }
10062
+
10063
+ declare interface StepStart {
10064
+ event_type: 'step.start';
10065
+ index: number;
10066
+ /**
10067
+ * A step in the interaction.
10068
+ */
10069
+ step: Step;
10070
+ /**
10071
+ * The event_id token to be used to resume the interaction stream, from this event.
10072
+ */
10073
+ event_id?: string;
10074
+ }
10075
+
10076
+ declare interface StepStop {
10077
+ event_type: 'step.stop';
10078
+ index: number;
10079
+ /**
10080
+ * The event_id token to be used to resume the interaction stream, from this event.
10081
+ */
10082
+ event_id?: string;
10083
+ }
10084
+
10101
10085
  declare class Stream<Item> implements AsyncIterable<Item> {
10102
10086
  private iterator;
10103
10087
  controller: AbortController;
@@ -10333,6 +10317,24 @@ declare interface TextContent {
10333
10317
  annotations?: Array<Annotation>;
10334
10318
  }
10335
10319
 
10320
+ /**
10321
+ * Configuration for text output format.
10322
+ */
10323
+ declare interface TextResponseFormat {
10324
+ type: 'text';
10325
+ /**
10326
+ * The MIME type of the text output.
10327
+ */
10328
+ mimeType?: 'application/json' | 'text/plain';
10329
+ /**
10330
+ * The JSON schema that the output should conform to. Only applicable when
10331
+ * mime_type is application/json.
10332
+ */
10333
+ schema?: {
10334
+ [key: string]: unknown;
10335
+ };
10336
+ }
10337
+
10336
10338
  /** The thinking features configuration. */
10337
10339
  export declare interface ThinkingConfig {
10338
10340
  /** Indicates whether to include thoughts in the response. If true, thoughts are returned only if the model supports thought and thoughts are available.
@@ -10372,12 +10374,12 @@ export declare enum ThinkingLevel {
10372
10374
  declare type ThinkingLevel_2 = 'minimal' | 'low' | 'medium' | 'high';
10373
10375
 
10374
10376
  /**
10375
- * A thought content block.
10377
+ * A thought step.
10376
10378
  */
10377
- declare interface ThoughtContent {
10379
+ declare interface ThoughtStep {
10378
10380
  type: 'thought';
10379
10381
  /**
10380
- * Signature to match the backend source to be part of the generation.
10382
+ * A signature hash for backend validation.
10381
10383
  */
10382
10384
  signature?: string;
10383
10385
  /**
@@ -11086,14 +11088,6 @@ export declare interface TuningValidationDataset {
11086
11088
  vertexDatasetResource?: string;
11087
11089
  }
11088
11090
 
11089
- declare interface Turn {
11090
- content?: Array<Content_2> | string;
11091
- /**
11092
- * The originator of this turn. Must be user for input or model for model output.
11093
- */
11094
- role?: string;
11095
- }
11096
-
11097
11091
  /** The reason why the turn is complete. */
11098
11092
  export declare enum TurnCompleteReason {
11099
11093
  /**
@@ -11760,6 +11754,7 @@ declare namespace types {
11760
11754
  EmbedContentParameters,
11761
11755
  UploadToFileSearchStoreResponse,
11762
11756
  UploadToFileSearchStoreOperation,
11757
+ DownloadMediaConfig,
11763
11758
  BlobImageUnion,
11764
11759
  PartUnion,
11765
11760
  PartListUnion,
@@ -12092,9 +12087,9 @@ declare interface URLContextCallArguments {
12092
12087
  }
12093
12088
 
12094
12089
  /**
12095
- * URL context content.
12090
+ * URL context call step.
12096
12091
  */
12097
- declare interface URLContextCallContent {
12092
+ declare interface URLContextCallStep {
12098
12093
  /**
12099
12094
  * Required. A unique ID for this specific tool call.
12100
12095
  */
@@ -12102,7 +12097,7 @@ declare interface URLContextCallContent {
12102
12097
  /**
12103
12098
  * Required. The arguments to pass to the URL context.
12104
12099
  */
12105
- arguments: URLContextCallArguments;
12100
+ arguments: URLContextCallStep.Arguments;
12106
12101
  type: 'url_context_call';
12107
12102
  /**
12108
12103
  * A signature hash for backend validation.
@@ -12110,6 +12105,18 @@ declare interface URLContextCallContent {
12110
12105
  signature?: string;
12111
12106
  }
12112
12107
 
12108
+ declare namespace URLContextCallStep {
12109
+ /**
12110
+ * Required. The arguments to pass to the URL context.
12111
+ */
12112
+ interface Arguments {
12113
+ /**
12114
+ * The URLs to fetch.
12115
+ */
12116
+ urls?: Array<string>;
12117
+ }
12118
+ }
12119
+
12113
12120
  /** Metadata returned when the model uses the `url_context` tool to get information from a user-provided URL. */
12114
12121
  export declare interface UrlContextMetadata {
12115
12122
  /** Output only. A list of URL metadata, with one entry for each URL retrieved by the tool. */
@@ -12131,9 +12138,9 @@ declare interface URLContextResult {
12131
12138
  }
12132
12139
 
12133
12140
  /**
12134
- * URL context result content.
12141
+ * URL context result step.
12135
12142
  */
12136
- declare interface URLContextResultContent {
12143
+ declare interface URLContextResultStep {
12137
12144
  /**
12138
12145
  * Required. ID to match the ID from the function call block.
12139
12146
  */
@@ -12141,7 +12148,7 @@ declare interface URLContextResultContent {
12141
12148
  /**
12142
12149
  * Required. The results of the URL context.
12143
12150
  */
12144
- result: Array<URLContextResult>;
12151
+ result: Array<URLContextResultStep.Result>;
12145
12152
  type: 'url_context_result';
12146
12153
  /**
12147
12154
  * Whether the URL context resulted in an error.
@@ -12153,6 +12160,22 @@ declare interface URLContextResultContent {
12153
12160
  signature?: string;
12154
12161
  }
12155
12162
 
12163
+ declare namespace URLContextResultStep {
12164
+ /**
12165
+ * The result of the URL context.
12166
+ */
12167
+ interface Result {
12168
+ /**
12169
+ * The status of the URL retrieval.
12170
+ */
12171
+ status?: 'success' | 'error' | 'paywall' | 'unsafe';
12172
+ /**
12173
+ * The URL that was fetched.
12174
+ */
12175
+ url?: string;
12176
+ }
12177
+ }
12178
+
12156
12179
  /** The metadata for a single URL retrieval. */
12157
12180
  export declare interface UrlMetadata {
12158
12181
  /** The URL retrieved by the tool. */
@@ -12331,6 +12354,14 @@ export declare interface UsageMetadata {
12331
12354
  trafficType?: TrafficType;
12332
12355
  }
12333
12356
 
12357
+ /**
12358
+ * Input provided by the user.
12359
+ */
12360
+ declare interface UserInputStep {
12361
+ type: 'user_input';
12362
+ content?: Array<Content_2>;
12363
+ }
12364
+
12334
12365
  /** The type of the VAD signal. */
12335
12366
  export declare enum VadSignalType {
12336
12367
  /**
@@ -12544,6 +12575,13 @@ export declare interface VideoMetadata {
12544
12575
  startOffset?: string;
12545
12576
  }
12546
12577
 
12578
+ /**
12579
+ * Configuration for video output format.
12580
+ */
12581
+ declare interface VideoResponseFormat {
12582
+ type: 'video';
12583
+ }
12584
+
12547
12585
  /** Voice activity signal. */
12548
12586
  export declare interface VoiceActivity {
12549
12587
  /** The type of the voice activity signal. */