@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.
package/dist/genai.d.ts CHANGED
@@ -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
  */
@@ -5480,9 +5241,9 @@ declare interface GoogleMapsCallArguments {
5480
5241
  }
5481
5242
 
5482
5243
  /**
5483
- * Google Maps content.
5244
+ * Google Maps call step.
5484
5245
  */
5485
- declare interface GoogleMapsCallContent {
5246
+ declare interface GoogleMapsCallStep {
5486
5247
  /**
5487
5248
  * Required. A unique ID for this specific tool call.
5488
5249
  */
@@ -5491,13 +5252,25 @@ declare interface GoogleMapsCallContent {
5491
5252
  /**
5492
5253
  * The arguments to pass to the Google Maps tool.
5493
5254
  */
5494
- arguments?: GoogleMapsCallArguments;
5255
+ arguments?: GoogleMapsCallStep.Arguments;
5495
5256
  /**
5496
5257
  * A signature hash for backend validation.
5497
5258
  */
5498
5259
  signature?: string;
5499
5260
  }
5500
5261
 
5262
+ declare namespace GoogleMapsCallStep {
5263
+ /**
5264
+ * The arguments to pass to the Google Maps tool.
5265
+ */
5266
+ interface Arguments {
5267
+ /**
5268
+ * The queries to be executed.
5269
+ */
5270
+ queries?: Array<string>;
5271
+ }
5272
+ }
5273
+
5501
5274
  /**
5502
5275
  * The result of the Google Maps.
5503
5276
  */
@@ -5555,17 +5328,14 @@ declare namespace GoogleMapsResult {
5555
5328
  }
5556
5329
 
5557
5330
  /**
5558
- * Google Maps result content.
5331
+ * Google Maps result step.
5559
5332
  */
5560
- declare interface GoogleMapsResultContent {
5333
+ declare interface GoogleMapsResultStep {
5561
5334
  /**
5562
5335
  * Required. ID to match the ID from the function call block.
5563
5336
  */
5564
5337
  call_id: string;
5565
- /**
5566
- * Required. The results of the Google Maps.
5567
- */
5568
- result: Array<GoogleMapsResult>;
5338
+ result: Array<GoogleMapsResultStep.Result>;
5569
5339
  type: 'google_maps_result';
5570
5340
  /**
5571
5341
  * A signature hash for backend validation.
@@ -5573,6 +5343,44 @@ declare interface GoogleMapsResultContent {
5573
5343
  signature?: string;
5574
5344
  }
5575
5345
 
5346
+ declare namespace GoogleMapsResultStep {
5347
+ /**
5348
+ * The result of the Google Maps.
5349
+ */
5350
+ interface Result {
5351
+ places?: Array<Result.Place>;
5352
+ widget_context_token?: string;
5353
+ }
5354
+ namespace Result {
5355
+ interface Place {
5356
+ name?: string;
5357
+ place_id?: string;
5358
+ review_snippets?: Array<Place.ReviewSnippet>;
5359
+ url?: string;
5360
+ }
5361
+ namespace Place {
5362
+ /**
5363
+ * Encapsulates a snippet of a user review that answers a question about the
5364
+ * features of a specific place in Google Maps.
5365
+ */
5366
+ interface ReviewSnippet {
5367
+ /**
5368
+ * The ID of the review snippet.
5369
+ */
5370
+ review_id?: string;
5371
+ /**
5372
+ * Title of the review.
5373
+ */
5374
+ title?: string;
5375
+ /**
5376
+ * A link that corresponds to the user review on Google Maps.
5377
+ */
5378
+ url?: string;
5379
+ }
5380
+ }
5381
+ }
5382
+ }
5383
+
5576
5384
  /** 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. */
5577
5385
  export declare interface GoogleRpcStatus {
5578
5386
  /** The status code, which should be an enum value of google.rpc.Code. */
@@ -5606,9 +5414,9 @@ declare interface GoogleSearchCallArguments {
5606
5414
  }
5607
5415
 
5608
5416
  /**
5609
- * Google Search content.
5417
+ * Google Search call step.
5610
5418
  */
5611
- declare interface GoogleSearchCallContent {
5419
+ declare interface GoogleSearchCallStep {
5612
5420
  /**
5613
5421
  * Required. A unique ID for this specific tool call.
5614
5422
  */
@@ -5616,7 +5424,7 @@ declare interface GoogleSearchCallContent {
5616
5424
  /**
5617
5425
  * Required. The arguments to pass to Google Search.
5618
5426
  */
5619
- arguments: GoogleSearchCallArguments;
5427
+ arguments: GoogleSearchCallStep.Arguments;
5620
5428
  type: 'google_search_call';
5621
5429
  /**
5622
5430
  * The type of search grounding enabled.
@@ -5628,6 +5436,18 @@ declare interface GoogleSearchCallContent {
5628
5436
  signature?: string;
5629
5437
  }
5630
5438
 
5439
+ declare namespace GoogleSearchCallStep {
5440
+ /**
5441
+ * Required. The arguments to pass to Google Search.
5442
+ */
5443
+ interface Arguments {
5444
+ /**
5445
+ * Web search queries for the following-up web search.
5446
+ */
5447
+ queries?: Array<string>;
5448
+ }
5449
+ }
5450
+
5631
5451
  /**
5632
5452
  * The result of the Google Search.
5633
5453
  */
@@ -5639,9 +5459,9 @@ declare interface GoogleSearchResult {
5639
5459
  }
5640
5460
 
5641
5461
  /**
5642
- * Google Search result content.
5462
+ * Google Search result step.
5643
5463
  */
5644
- declare interface GoogleSearchResultContent {
5464
+ declare interface GoogleSearchResultStep {
5645
5465
  /**
5646
5466
  * Required. ID to match the ID from the function call block.
5647
5467
  */
@@ -5649,7 +5469,7 @@ declare interface GoogleSearchResultContent {
5649
5469
  /**
5650
5470
  * Required. The results of the Google Search.
5651
5471
  */
5652
- result: Array<GoogleSearchResult>;
5472
+ result: Array<GoogleSearchResultStep.Result>;
5653
5473
  type: 'google_search_result';
5654
5474
  /**
5655
5475
  * Whether the Google Search resulted in an error.
@@ -5661,6 +5481,18 @@ declare interface GoogleSearchResultContent {
5661
5481
  signature?: string;
5662
5482
  }
5663
5483
 
5484
+ declare namespace GoogleSearchResultStep {
5485
+ /**
5486
+ * The result of the Google Search.
5487
+ */
5488
+ interface Result {
5489
+ /**
5490
+ * Web content snippet that can be embedded in a web page or an app webview.
5491
+ */
5492
+ search_suggestions?: string;
5493
+ }
5494
+ }
5495
+
5664
5496
  /** Tool to retrieve public web data for grounding, powered by Google. */
5665
5497
  export declare interface GoogleSearchRetrieval {
5666
5498
  /** Specifies the dynamic retrieval configuration for the given source. */
@@ -5814,6 +5646,8 @@ export declare interface GroundingChunkRetrievedContext {
5814
5646
  fileSearchStore?: string;
5815
5647
  /** Optional. Page number of the retrieved context. This field is not supported in Vertex AI. */
5816
5648
  pageNumber?: number;
5649
+ /** Optional. Media ID. This field is not supported in Vertex AI. */
5650
+ mediaId?: string;
5817
5651
  }
5818
5652
 
5819
5653
  /** A list of string values. This data type is not supported in Vertex AI. */
@@ -6276,6 +6110,29 @@ export declare enum ImageResizeMode {
6276
6110
  PAD = "PAD"
6277
6111
  }
6278
6112
 
6113
+ /**
6114
+ * Configuration for image output format.
6115
+ */
6116
+ declare interface ImageResponseFormat {
6117
+ type: 'image';
6118
+ /**
6119
+ * The aspect ratio for the image output.
6120
+ */
6121
+ 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';
6122
+ /**
6123
+ * The delivery mode for the image output.
6124
+ */
6125
+ delivery?: 'inline' | 'url';
6126
+ /**
6127
+ * The size of the image output.
6128
+ */
6129
+ imageSize?: '512' | '1K' | '2K' | '4K';
6130
+ /**
6131
+ * The MIME type of the image output.
6132
+ */
6133
+ mimeType?: 'image/jpeg';
6134
+ }
6135
+
6279
6136
  /** Image search for grounding and related configurations. */
6280
6137
  export declare interface ImageSearch {
6281
6138
  }
@@ -6407,15 +6264,11 @@ declare interface Interaction {
6407
6264
  /**
6408
6265
  * The input for the interaction.
6409
6266
  */
6410
- 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;
6267
+ input?: string | Array<Step> | Array<Content_2> | TextContent | ImageContent | AudioContent | DocumentContent | VideoContent;
6411
6268
  /**
6412
6269
  * The name of the `Model` used for generating the interaction.
6413
6270
  */
6414
6271
  model?: Model_2;
6415
- /**
6416
- * Output only. Responses from the model.
6417
- */
6418
- outputs?: Array<Content_2>;
6419
6272
  /**
6420
6273
  * The ID of the previous interaction, if any.
6421
6274
  */
@@ -6424,7 +6277,7 @@ declare interface Interaction {
6424
6277
  * Enforces that the generated response is a JSON object that complies with the
6425
6278
  * JSON schema specified in this field.
6426
6279
  */
6427
- response_format?: unknown;
6280
+ response_format?: Array<AudioResponseFormat | TextResponseFormat | ImageResponseFormat | VideoResponseFormat | unknown> | AudioResponseFormat | TextResponseFormat | ImageResponseFormat | VideoResponseFormat | unknown;
6428
6281
  /**
6429
6282
  * The mime type of the response. This is required if response_format is set.
6430
6283
  */
@@ -6434,13 +6287,17 @@ declare interface Interaction {
6434
6287
  */
6435
6288
  response_modalities?: Array<'text' | 'image' | 'audio' | 'video' | 'document'>;
6436
6289
  /**
6437
- * Output only. The role of the interaction.
6290
+ * @deprecated Output only. The role of the interaction.
6438
6291
  */
6439
6292
  role?: string;
6440
6293
  /**
6441
6294
  * The service tier for the interaction.
6442
6295
  */
6443
6296
  service_tier?: 'flex' | 'standard' | 'priority';
6297
+ /**
6298
+ * Output only. The steps that make up the interaction.
6299
+ */
6300
+ steps?: Array<Step>;
6444
6301
  /**
6445
6302
  * System instruction for the interaction.
6446
6303
  */
@@ -6467,8 +6324,8 @@ declare interface InteractionCancelParams {
6467
6324
  api_version?: string;
6468
6325
  }
6469
6326
 
6470
- declare interface InteractionCompleteEvent {
6471
- event_type: 'interaction.complete';
6327
+ declare interface InteractionCompletedEvent {
6328
+ event_type: 'interaction.completed';
6472
6329
  /**
6473
6330
  * Required. The completed interaction with empty outputs to reduce the payload
6474
6331
  * size. Use the preceding ContentDelta events for the actual output.
@@ -6480,6 +6337,18 @@ declare interface InteractionCompleteEvent {
6480
6337
  event_id?: string;
6481
6338
  }
6482
6339
 
6340
+ declare interface InteractionCreatedEvent {
6341
+ event_type: 'interaction.created';
6342
+ /**
6343
+ * The Interaction resource.
6344
+ */
6345
+ interaction: Interaction;
6346
+ /**
6347
+ * The event_id token to be used to resume the interaction stream, from this event.
6348
+ */
6349
+ event_id?: string;
6350
+ }
6351
+
6483
6352
  declare type InteractionCreateParams = CreateModelInteractionParamsNonStreaming | CreateModelInteractionParamsStreaming | CreateAgentInteractionParamsNonStreaming | CreateAgentInteractionParamsStreaming;
6484
6353
 
6485
6354
  declare interface InteractionDeleteParams {
@@ -6540,7 +6409,7 @@ export declare class Interactions extends BaseInteractions {
6540
6409
  }
6541
6410
 
6542
6411
  export declare namespace Interactions {
6543
- 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, };
6412
+ 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, };
6544
6413
  }
6545
6414
 
6546
6415
  declare namespace InteractionsAPI {
@@ -6550,58 +6419,64 @@ declare namespace InteractionsAPI {
6550
6419
  AllowedTools,
6551
6420
  Annotation,
6552
6421
  AudioContent,
6422
+ AudioResponseFormat,
6553
6423
  CodeExecutionCallArguments,
6554
- CodeExecutionCallContent,
6555
- CodeExecutionResultContent,
6424
+ CodeExecutionCallStep,
6425
+ CodeExecutionResultStep,
6556
6426
  Content_2 as Content,
6557
- ContentDelta,
6558
- ContentStart,
6559
- ContentStop,
6560
6427
  DeepResearchAgentConfig,
6561
6428
  DocumentContent,
6562
6429
  DynamicAgentConfig,
6563
6430
  ErrorEvent_2 as ErrorEvent,
6564
6431
  FileCitation,
6565
- FileSearchCallContent,
6566
- FileSearchResultContent,
6432
+ FileSearchCallStep,
6433
+ FileSearchResultStep,
6567
6434
  Function_2 as Function,
6568
- FunctionCallContent,
6569
- FunctionResultContent,
6435
+ FunctionCallStep,
6436
+ FunctionResultStep,
6570
6437
  GenerationConfig_2 as GenerationConfig,
6571
6438
  GoogleMapsCallArguments,
6572
- GoogleMapsCallContent,
6439
+ GoogleMapsCallStep,
6573
6440
  GoogleMapsResult,
6574
- GoogleMapsResultContent,
6441
+ GoogleMapsResultStep,
6575
6442
  GoogleSearchCallArguments,
6576
- GoogleSearchCallContent,
6443
+ GoogleSearchCallStep,
6577
6444
  GoogleSearchResult,
6578
- GoogleSearchResultContent,
6445
+ GoogleSearchResultStep,
6579
6446
  ImageConfig_2 as ImageConfig,
6580
6447
  ImageContent,
6448
+ ImageResponseFormat,
6581
6449
  Interaction,
6582
- InteractionCompleteEvent,
6450
+ InteractionCompletedEvent,
6451
+ InteractionCreatedEvent,
6583
6452
  InteractionSSEEvent,
6584
- InteractionStartEvent,
6585
6453
  InteractionStatusUpdate,
6586
- MCPServerToolCallContent,
6587
- MCPServerToolResultContent,
6454
+ MCPServerToolCallStep,
6455
+ MCPServerToolResultStep,
6588
6456
  Model_2 as Model,
6457
+ ModelOutputStep,
6589
6458
  PlaceCitation,
6590
6459
  SpeechConfig_2 as SpeechConfig,
6460
+ Step,
6461
+ StepDelta,
6462
+ StepStart,
6463
+ StepStop,
6591
6464
  TextContent,
6465
+ TextResponseFormat,
6592
6466
  ThinkingLevel_2 as ThinkingLevel,
6593
- ThoughtContent,
6467
+ ThoughtStep,
6594
6468
  Tool_2 as Tool,
6595
6469
  ToolChoiceConfig,
6596
6470
  ToolChoiceType,
6597
- Turn,
6598
6471
  URLCitation,
6599
6472
  URLContextCallArguments,
6600
- URLContextCallContent,
6473
+ URLContextCallStep,
6601
6474
  URLContextResult,
6602
- URLContextResultContent,
6475
+ URLContextResultStep,
6603
6476
  Usage,
6477
+ UserInputStep,
6604
6478
  VideoContent,
6479
+ VideoResponseFormat,
6605
6480
  WebhookConfig_2 as WebhookConfig,
6606
6481
  InteractionDeleteResponse,
6607
6482
  InteractionCreateParams,
@@ -6620,19 +6495,7 @@ declare namespace InteractionsAPI {
6620
6495
  }
6621
6496
  }
6622
6497
 
6623
- declare type InteractionSSEEvent = InteractionStartEvent | InteractionCompleteEvent | InteractionStatusUpdate | ContentStart | ContentDelta | ContentStop | ErrorEvent_2;
6624
-
6625
- declare interface InteractionStartEvent {
6626
- event_type: 'interaction.start';
6627
- /**
6628
- * The Interaction resource.
6629
- */
6630
- interaction: Interaction;
6631
- /**
6632
- * The event_id token to be used to resume the interaction stream, from this event.
6633
- */
6634
- event_id?: string;
6635
- }
6498
+ declare type InteractionSSEEvent = InteractionCreatedEvent | InteractionCompletedEvent | InteractionStatusUpdate | ErrorEvent_2 | StepStart | StepDelta | StepStop;
6636
6499
 
6637
6500
  declare interface InteractionStatusUpdate {
6638
6501
  event_type: 'interaction.status_update';
@@ -7835,9 +7698,9 @@ export declare interface McpServer {
7835
7698
  }
7836
7699
 
7837
7700
  /**
7838
- * MCPServer tool call content.
7701
+ * MCPServer tool call step.
7839
7702
  */
7840
- declare interface MCPServerToolCallContent {
7703
+ declare interface MCPServerToolCallStep {
7841
7704
  /**
7842
7705
  * Required. A unique ID for this specific tool call.
7843
7706
  */
@@ -7864,9 +7727,9 @@ declare interface MCPServerToolCallContent {
7864
7727
  }
7865
7728
 
7866
7729
  /**
7867
- * MCPServer tool result content.
7730
+ * MCPServer tool result step.
7868
7731
  */
7869
- declare interface MCPServerToolResultContent {
7732
+ declare interface MCPServerToolResultStep {
7870
7733
  /**
7871
7734
  * Required. ID to match the ID from the function call block.
7872
7735
  */
@@ -7874,7 +7737,7 @@ declare interface MCPServerToolResultContent {
7874
7737
  /**
7875
7738
  * The output from the MCP server call. Can be simple text or rich content.
7876
7739
  */
7877
- result: unknown | Array<TextContent | ImageContent> | string;
7740
+ result: unknown | string | Array<TextContent | ImageContent>;
7878
7741
  type: 'mcp_server_tool_result';
7879
7742
  /**
7880
7743
  * Name of the tool which is called for this specific tool call.
@@ -8059,6 +7922,14 @@ export declare interface ModelArmorConfig {
8059
7922
  responseTemplateName?: string;
8060
7923
  }
8061
7924
 
7925
+ /**
7926
+ * Output generated by the model.
7927
+ */
7928
+ declare interface ModelOutputStep {
7929
+ type: 'model_output';
7930
+ content?: Array<Content_2>;
7931
+ }
7932
+
8062
7933
  export declare class Models extends BaseModule {
8063
7934
  private readonly apiClient;
8064
7935
  constructor(apiClient: ApiClient);
@@ -10086,6 +9957,119 @@ export declare enum StartSensitivity {
10086
9957
  START_SENSITIVITY_LOW = "START_SENSITIVITY_LOW"
10087
9958
  }
10088
9959
 
9960
+ /**
9961
+ * A step in the interaction.
9962
+ */
9963
+ declare type Step = UserInputStep | ModelOutputStep | ThoughtStep | FunctionCallStep | CodeExecutionCallStep | URLContextCallStep | MCPServerToolCallStep | GoogleSearchCallStep | FileSearchCallStep | GoogleMapsCallStep | FunctionResultStep | CodeExecutionResultStep | URLContextResultStep | GoogleSearchResultStep | MCPServerToolResultStep | FileSearchResultStep | GoogleMapsResultStep;
9964
+
9965
+ declare interface StepDelta {
9966
+ delta: StepDelta.Text | StepDelta.Image | StepDelta.Audio | StepDelta.Document | StepDelta.Video | StepDelta.ThoughtSummary | StepDelta.ThoughtSignature | StepDelta.TextAnnotationDelta | StepDelta.ArgumentsDelta;
9967
+ event_type: 'step.delta';
9968
+ index: number;
9969
+ /**
9970
+ * The event_id token to be used to resume the interaction stream, from this event.
9971
+ */
9972
+ event_id?: string;
9973
+ }
9974
+
9975
+ declare namespace StepDelta {
9976
+ interface Text {
9977
+ text: string;
9978
+ type: 'text';
9979
+ }
9980
+ interface Image {
9981
+ type: 'image';
9982
+ data?: string;
9983
+ mime_type?: 'image/png' | 'image/jpeg' | 'image/webp' | 'image/heic' | 'image/heif' | 'image/gif' | 'image/bmp' | 'image/tiff';
9984
+ /**
9985
+ * The resolution of the media.
9986
+ */
9987
+ resolution?: 'low' | 'medium' | 'high' | 'ultra_high';
9988
+ uri?: string;
9989
+ }
9990
+ interface Audio {
9991
+ type: 'audio';
9992
+ /**
9993
+ * The number of audio channels.
9994
+ */
9995
+ channels?: number;
9996
+ data?: string;
9997
+ 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';
9998
+ /**
9999
+ * @deprecated Deprecated. Use sample_rate instead. The value is ignored.
10000
+ */
10001
+ rate?: number;
10002
+ /**
10003
+ * The sample rate of the audio.
10004
+ */
10005
+ sample_rate?: number;
10006
+ uri?: string;
10007
+ }
10008
+ interface Document {
10009
+ type: 'document';
10010
+ data?: string;
10011
+ mime_type?: 'application/pdf';
10012
+ uri?: string;
10013
+ }
10014
+ interface Video {
10015
+ type: 'video';
10016
+ data?: string;
10017
+ mime_type?: 'video/mp4' | 'video/mpeg' | 'video/mpg' | 'video/mov' | 'video/avi' | 'video/x-flv' | 'video/webm' | 'video/wmv' | 'video/3gpp';
10018
+ /**
10019
+ * The resolution of the media.
10020
+ */
10021
+ resolution?: 'low' | 'medium' | 'high' | 'ultra_high';
10022
+ uri?: string;
10023
+ }
10024
+ interface ThoughtSummary {
10025
+ type: 'thought_summary';
10026
+ /**
10027
+ * A new summary item to be added to the thought.
10028
+ */
10029
+ content?: InteractionsAPI.TextContent | InteractionsAPI.ImageContent;
10030
+ }
10031
+ interface ThoughtSignature {
10032
+ type: 'thought_signature';
10033
+ /**
10034
+ * Signature to match the backend source to be part of the generation.
10035
+ */
10036
+ signature?: string;
10037
+ }
10038
+ interface TextAnnotationDelta {
10039
+ type: 'text_annotation_delta';
10040
+ /**
10041
+ * Citation information for model-generated content.
10042
+ */
10043
+ annotations?: Array<InteractionsAPI.Annotation>;
10044
+ }
10045
+ interface ArgumentsDelta {
10046
+ type: 'arguments_delta';
10047
+ partial_arguments?: string;
10048
+ }
10049
+ }
10050
+
10051
+ declare interface StepStart {
10052
+ event_type: 'step.start';
10053
+ index: number;
10054
+ /**
10055
+ * A step in the interaction.
10056
+ */
10057
+ step: Step;
10058
+ /**
10059
+ * The event_id token to be used to resume the interaction stream, from this event.
10060
+ */
10061
+ event_id?: string;
10062
+ }
10063
+
10064
+ declare interface StepStop {
10065
+ event_type: 'step.stop';
10066
+ index: number;
10067
+ /**
10068
+ * The event_id token to be used to resume the interaction stream, from this event.
10069
+ */
10070
+ event_id?: string;
10071
+ }
10072
+
10089
10073
  declare class Stream<Item> implements AsyncIterable<Item> {
10090
10074
  private iterator;
10091
10075
  controller: AbortController;
@@ -10321,6 +10305,24 @@ declare interface TextContent {
10321
10305
  annotations?: Array<Annotation>;
10322
10306
  }
10323
10307
 
10308
+ /**
10309
+ * Configuration for text output format.
10310
+ */
10311
+ declare interface TextResponseFormat {
10312
+ type: 'text';
10313
+ /**
10314
+ * The MIME type of the text output.
10315
+ */
10316
+ mimeType?: 'application/json' | 'text/plain';
10317
+ /**
10318
+ * The JSON schema that the output should conform to. Only applicable when
10319
+ * mime_type is application/json.
10320
+ */
10321
+ schema?: {
10322
+ [key: string]: unknown;
10323
+ };
10324
+ }
10325
+
10324
10326
  /** The thinking features configuration. */
10325
10327
  export declare interface ThinkingConfig {
10326
10328
  /** Indicates whether to include thoughts in the response. If true, thoughts are returned only if the model supports thought and thoughts are available.
@@ -10360,12 +10362,12 @@ export declare enum ThinkingLevel {
10360
10362
  declare type ThinkingLevel_2 = 'minimal' | 'low' | 'medium' | 'high';
10361
10363
 
10362
10364
  /**
10363
- * A thought content block.
10365
+ * A thought step.
10364
10366
  */
10365
- declare interface ThoughtContent {
10367
+ declare interface ThoughtStep {
10366
10368
  type: 'thought';
10367
10369
  /**
10368
- * Signature to match the backend source to be part of the generation.
10370
+ * A signature hash for backend validation.
10369
10371
  */
10370
10372
  signature?: string;
10371
10373
  /**
@@ -11074,14 +11076,6 @@ export declare interface TuningValidationDataset {
11074
11076
  vertexDatasetResource?: string;
11075
11077
  }
11076
11078
 
11077
- declare interface Turn {
11078
- content?: Array<Content_2> | string;
11079
- /**
11080
- * The originator of this turn. Must be user for input or model for model output.
11081
- */
11082
- role?: string;
11083
- }
11084
-
11085
11079
  /** The reason why the turn is complete. */
11086
11080
  export declare enum TurnCompleteReason {
11087
11081
  /**
@@ -11748,6 +11742,7 @@ declare namespace types {
11748
11742
  EmbedContentParameters,
11749
11743
  UploadToFileSearchStoreResponse,
11750
11744
  UploadToFileSearchStoreOperation,
11745
+ DownloadMediaConfig,
11751
11746
  BlobImageUnion,
11752
11747
  PartUnion,
11753
11748
  PartListUnion,
@@ -12080,9 +12075,9 @@ declare interface URLContextCallArguments {
12080
12075
  }
12081
12076
 
12082
12077
  /**
12083
- * URL context content.
12078
+ * URL context call step.
12084
12079
  */
12085
- declare interface URLContextCallContent {
12080
+ declare interface URLContextCallStep {
12086
12081
  /**
12087
12082
  * Required. A unique ID for this specific tool call.
12088
12083
  */
@@ -12090,7 +12085,7 @@ declare interface URLContextCallContent {
12090
12085
  /**
12091
12086
  * Required. The arguments to pass to the URL context.
12092
12087
  */
12093
- arguments: URLContextCallArguments;
12088
+ arguments: URLContextCallStep.Arguments;
12094
12089
  type: 'url_context_call';
12095
12090
  /**
12096
12091
  * A signature hash for backend validation.
@@ -12098,6 +12093,18 @@ declare interface URLContextCallContent {
12098
12093
  signature?: string;
12099
12094
  }
12100
12095
 
12096
+ declare namespace URLContextCallStep {
12097
+ /**
12098
+ * Required. The arguments to pass to the URL context.
12099
+ */
12100
+ interface Arguments {
12101
+ /**
12102
+ * The URLs to fetch.
12103
+ */
12104
+ urls?: Array<string>;
12105
+ }
12106
+ }
12107
+
12101
12108
  /** Metadata returned when the model uses the `url_context` tool to get information from a user-provided URL. */
12102
12109
  export declare interface UrlContextMetadata {
12103
12110
  /** Output only. A list of URL metadata, with one entry for each URL retrieved by the tool. */
@@ -12119,9 +12126,9 @@ declare interface URLContextResult {
12119
12126
  }
12120
12127
 
12121
12128
  /**
12122
- * URL context result content.
12129
+ * URL context result step.
12123
12130
  */
12124
- declare interface URLContextResultContent {
12131
+ declare interface URLContextResultStep {
12125
12132
  /**
12126
12133
  * Required. ID to match the ID from the function call block.
12127
12134
  */
@@ -12129,7 +12136,7 @@ declare interface URLContextResultContent {
12129
12136
  /**
12130
12137
  * Required. The results of the URL context.
12131
12138
  */
12132
- result: Array<URLContextResult>;
12139
+ result: Array<URLContextResultStep.Result>;
12133
12140
  type: 'url_context_result';
12134
12141
  /**
12135
12142
  * Whether the URL context resulted in an error.
@@ -12141,6 +12148,22 @@ declare interface URLContextResultContent {
12141
12148
  signature?: string;
12142
12149
  }
12143
12150
 
12151
+ declare namespace URLContextResultStep {
12152
+ /**
12153
+ * The result of the URL context.
12154
+ */
12155
+ interface Result {
12156
+ /**
12157
+ * The status of the URL retrieval.
12158
+ */
12159
+ status?: 'success' | 'error' | 'paywall' | 'unsafe';
12160
+ /**
12161
+ * The URL that was fetched.
12162
+ */
12163
+ url?: string;
12164
+ }
12165
+ }
12166
+
12144
12167
  /** The metadata for a single URL retrieval. */
12145
12168
  export declare interface UrlMetadata {
12146
12169
  /** The URL retrieved by the tool. */
@@ -12319,6 +12342,14 @@ export declare interface UsageMetadata {
12319
12342
  trafficType?: TrafficType;
12320
12343
  }
12321
12344
 
12345
+ /**
12346
+ * Input provided by the user.
12347
+ */
12348
+ declare interface UserInputStep {
12349
+ type: 'user_input';
12350
+ content?: Array<Content_2>;
12351
+ }
12352
+
12322
12353
  /** The type of the VAD signal. */
12323
12354
  export declare enum VadSignalType {
12324
12355
  /**
@@ -12532,6 +12563,13 @@ export declare interface VideoMetadata {
12532
12563
  startOffset?: string;
12533
12564
  }
12534
12565
 
12566
+ /**
12567
+ * Configuration for video output format.
12568
+ */
12569
+ declare interface VideoResponseFormat {
12570
+ type: 'video';
12571
+ }
12572
+
12535
12573
  /** Voice activity signal. */
12536
12574
  export declare interface VoiceActivity {
12537
12575
  /** The type of the voice activity signal. */