1.48.0 (2026-03-31)
- Support dedicated TextAnnotationDelta for streaming tool responses (89552ba)
- Fix service_tier enums. (9bdc2ae)
1.47.0 (2026-03-27)
- Add custom_metadata to FileSearchResult. (083a1e3)
- Add labels field to Veo configs (930c9c3)
- Add mime type for Audio content (1ad80c6)
- Add model_status to GenerateContentResponse (Gemini API only) (5e1110d)
- Add part_metadata in Part (Gemini API only) (5e1110d)
- Add service tier for interactions. (406de38)
- Add service tier to GenerateContent. (0bfe800)
- Add support for more image and audio MIME types in Interactions content (baadbfd)
- Add supported models to the ModelOptions (94642b6)
- genai: add TURN_INCLUDES_AUDIO_ACTIVITY_AND_ALL_VIDEO to TurnCoverage (fdacac2)
- support hyperparameters in distillation tuning (716e021)
- Support rendered_parts in GroundingSupport (5e1110d)
- support us region routing (d391cff)
1.46.0 (2026-03-17)
- [Interactions] Breaking change to Interactions API to refactor TextContent annotations to use specific citation types (9fa8b1d)
- [Interactions] Breaking change for Interactions, rename ContentDelta unions. (917f24f)
- [Interactions] Breaking change to Interactions API to rename rendered_content to search_suggestions (cc6bd38)
- [Interactions] Add and update 'signature' fields for tool call/result content types. (e73ca5b)
- [Interactions] Support Google Maps in Interactions (d0593e3)
- Support include_server_side_tool_invocations for genai. (c627d6f)
- Quote
functionResponseskey in LiveClientMessage (9740426)
1.45.0 (2026-03-12)
- Add inference_generation_config to EvaluationConfig for Tuning (b4ac722)
- enable language code for audio transcription config in Live API for Vertex AI (67f39ec)
1.44.0 (2026-03-04)
- Add gemini-3.1-flash-image-preview model (175603b)
- Support signature for all Interaction tool types (b79108f)
- Update data types from discovery doc. (249b15a)
1.43.0 (2026-02-26)
- Add gemini-3.1-pro-preview to list of models in Interactions (14775fe)
- Add Image Grounding support to GoogleSearch tool (9187ca7)
- enable server side MCP and disable all other AFC when server side MCP is configured. (c7888c4)
- Support more image sizes and resolutions (54f4145)
- change interactions media mime type to enum (breaking change for experimental feature) (7c2d5b2)
1.42.0 (2026-02-18)
- Support multimodal embedding for Gemini Embedding 2.0 and support MaaS models in Models.embed_content() (Vertex AI API) (944ae97)
1.41.0 (2026-02-10)
- Add HTTP retry support to the SDK (37d4f2e)
- Add INCOMPLETE status to Interaction. (519f837)
- Pass retry options to the underlying Interactions client (086cf82)
- Support encryption_spec in tuning job creation configuration for GenAI SDK (e218add)
1.40.0 (2026-02-04)
- Update data types from discovery doc. (bad81a7)
1.39.0 (2026-01-30)
- Add
include_inputquery parameter to Get Interaction endpoint. (84bc7f4) - Add registerFiles for you can use gcs files with mldev. (7acf3e0)
- Support distillation tuning (c9a2e72)
- Support OSS Tuning in GenAI SDK (258f211)
- Add metadata in batch inlined response (0e95bb7)
1.38.0 (2026-01-20)
- Add ModelArmorConfig support for prompt and response sanitization via the Model Armor service (be65c24)
- Allow custom endpoints for authentication with Vertex AI in Typescript (891e32d), closes #1137
- Update data types from discovery doc. (a16a180)
- Update data types from discovery doc. (a4eaa8c)
1.37.0 (2026-01-15)
- Support 4:5 and 5:4 aspect ratio in Interactions (5294da2)
1.36.0 (2026-01-14)
- Add FileSearchCallContent to Interactions (ccf11a4)
- Add ImageConfig to GenerationConfig for image generation in Interactions (f34b6fb)
- Support JS local tokenizer (bb429ad)
- voice activity support (3ad1011)
1.35.0 (2026-01-07)
- [Python] add RegisterFiles so gcs files can be used with genai. (1bf555c)
- Add PersonGeneration to ImageConfig for Vertex Gempix (b0753d7)
1.34.0 (2025-12-16)
- Add minimal and medium thinking levels. (19f8969)
- Add support for Struct in ToolResult Content. (750a4f7)
- Add ultra high resolution to the media resolution in Parts. (fa9a99e)
- Add ULTRA_HIGH MediaResolution and new ThinkingLevel enums (b8a7d8c)
- Define and use DocumentMimeType for DocumentContent (c2f118d)
- support multi speaker for Vertex AI (94d681b)
1.33.0 (2025-12-11)
- Add IMAGE_RECITATION and IMAGE_OTHER enum values to FinishReason (ab87f51)
- Add voice activity detection signal. (bff9695)
- Make the Interactions API public. (a8dd67b)
1.32.0 (2025-12-09)
- Add enableEnhancedCivicAnswers feature in GenerateContentConfig (b429389)
- Support ReplicatedVoiceConfig (7a56e35)
- Improve stream parsing performance (c9c1359)
1.31.0 (2025-12-03)
- Add empty response for tunings.cancel() (5dc25bc)
1.30.0 (2025-11-18)
- add display name to FunctionResponseBlob (81036ed)
- add display name to FunctionResponseFileData (564f718)
- Add generate_content_config.thinking_level (4e586df)
- Add image output options to ImageConfig for Vertex (fff5fb6)
- Add part.media_resolution (4e586df)
- support Function call argument streaming for all languages (b3b94f6)
1.29.1 (2025-11-12)
- Add missing fields to the model types (8937903)
- Don't generate warnings from response.text property because of thought_signature. (356b843)
- Fix base_steps parameter for recontext_image (946415f)
1.29.0 (2025-11-05)
- add complete stats to BatchJob (3263514)
- Add FileSearch tool and associated FileSearchStore management APIs (05cff8c)
- Add FileSearch tool and associated FileSearchStore management APIs (25aa2a4)
- Add image_size to ImageConfig (Early Access Program) (e1e883a)
- Added phish filtering feature. (4b75e5d)
- update the build configs (97a2fac)
- disable AFC when there are AFC incompatible tool presented. (30c63b4)
1.28.0 (2025-10-29)
- Add safety_filter_level and person_generation for Imagen upscaling (aaceb1b)
- Add support for preference optimization tuning in the SDK. (f74660e)
- Pass file name to the backend when uploading with a file path (7115726)
- support default global location when not using api key with vertexai backend (f8967b1)
- Update
test-server-sdkdependency to^0.2.9. (8d507bf)
- Fix downloader. With this change after you await the download, the file write is complete, and the data is fully readable. (127c9bf)
1.27.0 (2025-10-23)
- Remove redundant calls to rollup/api-extractor, speeds up the TS SDK testing further by a factor of 2. (009b75f)
- google auth ADC error (f87611c)
1.26.0 (2025-10-21)
- Add enable_enhanced_civic_answers in GenerationConfig (0e95d72)
- support jailbreak in HarmCategory and BlockedReason (cd79a83)
1.25.0 (2025-10-15)
- Support video extension for Veo on Gemini Developer API (5d11d81)
1.24.0 (2025-10-10)
- Enable Google Maps tool for Genai. (b671fe9)
- Support enableWidget feature in GoogleMaps (81ed6a3)
- Support Gemini batch inline request's metadata and add test coverage to safety setting (722562b)
1.23.0 (2025-10-08)
- Add
NO_IMAGEenum value toFinishReason(5305334) - Add labels field to Imagen configs (1e9e96c)
- Add utility methods for creating
FunctionResponsePartand creating FunctionResponsePartwithFunctionResponseParts(5a0ce38) - Enable Ingredients to Video and Advanced Controls for Veo on Gemini Developer API (Early Access Program) (d5bd0c1)
- support system_instruction and tools mapping via converter and cleanup handwritten mapping (54ec204)
1.22.0 (2025-10-02)
- Add
ImageConfigtoGenerateContentConfig(c137172) - Add thinking_config for live (decc090)
- rename ComputerUse tool (early access) (e4345f9)
1.21.0 (2025-09-25)
- Add FunctionResponsePart & ToolComputerUse.excludedPredefinedFunctions (61620f6)
- Support Imagen 4 Ingredients on Vertex (56cccfc)
- #963 allow leading whitespace before data chunks (24f4153)
- Expose
JOB_STATE_RUNNINGandJOB_STATE_EXPIREDfor Gemini Batches states (b513c69) - initialization of
pre_tuned_model_checkpoint_idfrom tuning config. (2e10e7f)
1.20.0 (2025-09-16)
- Add 'turn_complete_reason' and 'waiting_for_input' fields. (cb54332)
- Add
VideoGenerationMaskModeenum for Veo 2 Editing (3e850fd)
1.19.0 (2025-09-09)
- Enable
idfield inFunctionCallfor Vertex AI. (7c6e047)
1.18.0 (2025-09-09)
1.17.0 (2025-09-02)
- Add resolution field for Gemini Developer API Veo 3 generation (5ce467b)
1.16.0 (2025-08-27)
- add
sdkHttpResponse.headersto *Delete responses. (397c602) - Add add_watermark field for recontext_image (Virtual Try-On, Product Recontext) (e182fe8)
- Add output_gcs_uri to Imagen upscale_image (8db102a)
- add outputSchema support for MCP (e2d4e50)
- Add VALIDATED mode into FunctionCallingConfigMode (af410e7)
- Add VideoGenerationReferenceType enum for generate_videos (eb17013)
- Support GenerateVideosSource for Veo GenerateVideos (2241749)
- support tunings.cancel in the genai SDK for Python, Java, JS, and Go (2d67d25)
- Update server test recordings to JSON. (dd0a033)
- js: empty env vars should be treated as undefined. (16a80c0)
1.15.0 (2025-08-18)
- Support Imagen image segmentation on Vertex (6b86bae)
- Support Veo 2 Reference Images to Video Generation on Vertex (6bca6ad)
- chats: Drop the validation on part with empty text (d2602d2)
- Fix the bug to support Gemini Batch inlined requests system instruction (843c774)
1.14.0 (2025-08-13)
- add an _exclude_fields characteristic in the config for unused values to avoid unnecessary and unused code in converter and type generation. (15a4504)
- enable continuous fine-tuning on a pre-tuned model in the SDK. (027f09d)
- support document name in grounding metadata (b801236)
- Support exclude_domains in Google Search and Enterprise Web Search (8ed49f1)
- Replace
gemini-2.5-flash-preview-04-17withgemini-2.5-flash. (d7f1c30)
1.13.0 (2025-08-06)
- Add image_size field for Gemini Developer API Imagen 4 generation (9841ecb)
- Add Lyria enum for music generation mode for vocalization (b05821f)
- allow methods in batch to return headers in sdk_http_response by default (55a2be4)
- enable responseId for Gemini Developer API (40e3a58)
- Support image recontext on Vertex (e417867)
- Support new enum types for UrlRetrievalStatus (88b47e6)
1.12.0 (2025-07-30)
- allow methods in models to return headers in sdk_http_response by default. (efdcb8a)
- allow methods in tuning to return headers in sdk_http_response by default (f81c16a)
- Correctly populated the Rai filter reasons on Video Generation (316cdb0)
1.11.0 (2025-07-23)
- Add image_size field for Vertex Imagen 4 generation (b6cc23e)
- enable function calling for Callable Tool besides McpClient in generateContent. (90565bd)
- Introduced Operations.get() which is a generic function to handle all Operation types. (a6f85c1)
- return headers for list method in all modules. (e8591a3)
- Updates to codegen instructions to improve performance for TypeScript (9c97ec6)
1.10.0 (2025-07-17)
- Add
addWatermarkparameter to the edit image configuration. (16bca99) - Adding codegen instructions to guide LLMs to generate code with the Google GenAI SDK (7f74175)
- Update
hasMcpToolUsageto set the MCP tool usage when it is detected by callingmcpToTool(a70f9c4)
- live: enhance security by moving api key from query parameters to header (f183b7d)
- live: Rollback commit f183b7d, breaks live in web env (be24e86)
1.9.0 (2025-07-09)
- Add new languages for Imagen 4 prompt language (65be769)
- add unref() for request timeouts (6b76c4b)
1.8.0 (2025-07-01)
- Enable Vertex Multimodal Dataset as input to supervised fine-tuning. (055b435)
- Support Batches delete (6cd9174)
- Support different media input in Vertex Live API (305952e)
1.7.0 (2025-06-25)
- Add compressionQuality enum for generate_videos (4325773)
- Add enhance_input_image and image_preservation_factor fields for upscale_image (13eedda)
- Batches support in TS (dd7355e)
- expose the responseJsonSchema in GenerateContentConfig (db54a5f)
1.6.0 (2025-06-21)
- allow users to access headers for generateContent method and generateContentStream (f5f0e02)
- enable json schema for controlled output and function declaration. (7d53d57)
- Include status code and export HTTP errors (bcabcb6)
- support extra_body in HttpOptions (1d48b6e)
- Fix build errors when the mcp package is not installed. (5b7e695)
1.5.1 (2025-06-13)
- Update MCP dependencies to be type-only (648019b)
1.5.0 (2025-06-11)
- add a timeout parameter to CallableToolConfig (06f31fd)
- Add datastore_spec field for VertexAISearch (1b18a52)
- Add support for Veo frame interpolation and video extension (53f22e6)
- Make the MCP SDK an optional peer dependency. (2472fd3)
- RAG - Introducing context storing for Gemini Live API. (bdc2355)
- Prioritize credentials over implicit api key (from env) for node client using Vertex AI (d82aba2), closes #616
1.4.0 (2025-06-04)
- Add enhance_prompt field for Gemini Developer API generate_videos (ea3cc8e)
- Enable url_context for Vertex (e377f58)
- js: Support
GEMINI_API_KEYas environment variable for setting API key. (97850ad)
- defined Type becomes TYPE_UNSPECIFIED when obfuscation is enabled (1ddf4f5)
- Update live SDK sample to address choppy audio quality (319bcbb)
- use Enum's value instead of keys for obfuscation. (132bd04)
1.3.0 (2025-05-30)
- Adding
thought_signaturefield to thePartto store the signature for thoughts. (20815b2) - include UNEXPECTED_TOOL_CALL enum value to FinishReason for Vertex AI APIs. (bd5a2bf)
- Support ephemeral auth tokens as API keys for live connections in TS. (507bfb5)
- Support ephemeral token creation in TS (425cfe6)
- Rename LiveEphemeralParameters to LiveConnectConstraints. (86e9652)
1.2.0 (2025-05-28)
- Remove unsupported Lyria enum for music generation mode
- Add generate_audio field for private testing of video generation (37e14c5)
- fix README typo (04259ad)
- Remove unsupported Lyria enum for music generation mode (0b935cd)
1.1.0 (2025-05-26)
- Add CommonJS modules (f40c47c)
- support new fields in FileData, GenerationConfig, GroundingChunkRetrievedContext, RetrievalConfig, Schema, TuningJob, VertexAISearch, (cd04548)
1.0.1 (2025-05-22)
Note
This version drops support for end-of-life Node.js v18.
- After an error on sendMessage, all subsequent calls fail with the same error (778abcc)
- Fixed sendMessage subsequent calls test to follow the arrange-act-assert pattern (778abcc)
- Unbreak direct
requires from node. (023efd5)
1.0.0 (2025-05-19)
- Fix Lyria method name for JS, update parameters type
- Support ephemeral token creation in Python (8e12730)
- Support using ephemeral token in Live session connection in Python (8e12730)
- allow McpClient to be passed in with AFC disabled. (b13f1f8)
- Blob handling in realtime music (f760755)
- Prevent MCP label from being appended multiple times if they already exist (c59ffe7)
- No longer preview. (5e137d4)
- Fix Lyria method name for JS, update parameters type (99dba6e)
0.15.0 (2025-05-19)
- add
time range filterto Google Search Tool (a938111) - Add basic support for async function calling. (8e2f357)
- Add CallableToolConfig to specify behavior for FunctionDeclarations (c4148d0)
- add configurations for afc. (ce7855b)
- add live proactivity_audio and enable_affective_dialog (20e3892)
- Add Lyria Realtime music generation support for JS (aefcaa5)
- Add Lyria Realtime Music Types (99255d8)
- Add MCP telemetry usage to TS SDK. (09a83e9)
- add multi-speaker voice config (6fe6074)
- Add support for lat/long in search. (4cd79f6)
- Add support for MCP in TS SDK. (921a4be)
- add support for propertyOrdering field. (a77822b)
- Add Video FPS, and enable start/end_offset for MLDev (19f20e9)
- Enable AFC on Generate content stream (ff2ce35)
- export the createJsonSchemaValidator methods to be available for other library. (b3359a1)
- List all mcp tools to max and mcpToTool taking a spread. (44cd9c9)
- MCP: Add a new interface for automatic function calling (dc49ffc)
- MCP: Add mcpToTool to pass MCP clients for automatic function calling (825f385)
- Remove MCP Tool and MCP Client from ToolUnion (d35e16d)
- support customer-managed encryption key in cached content (3e7437a)
- Support Url Context Retrieval tool (aaaf9a9)
- Add an ES module for node environments. (ff4bbd1)
- Add default headers to model calls when MCP is enabled (9442eea)
- Allow contents with no-text thoughts in chat history (4112d1c)
- chats: Enforce internal management of chat history (abe592f)
- chats: Relax the constraint on chat turns (68115a8)
- make the system-test build. (dd7154c)
- move test-server-sdk from dependencies to devDependencies (233a909)
- Move test-server-sdk to devDependencies (#574) (b64deeb)
- Run tests against Node 24 (28a56ac)
- Skip MCP tool call when function name not present in the tool (9f3d360)
0.14.1 (2025-05-15)
0.14.0 (2025-05-13)
- Add Imagen edit_image support in JS SDK (6c99936)
- Add Imagen upscale_image support for JS (6fe1a68)
- add support for audio, video, text and session resumption in java. (e5542c6)
- support display_name for Blob class when calling Vertex AI (fc35f51)
- Support tuning checkpoints (6bd9c9e)
0.13.0 (2025-05-07)
- Add
textanddataaccessors to LiveServerMessage (a3c4201) - Add
Tool.enterprise_web_searchfield (29b097d) - Add a models.list function to list available models. (477d9ec)
- Add Files.download method. (8f44c99)
- Add support for Grounding with Google Maps (41f0cc2)
- enable input transcription for Gemini API. (767c5d5)
- Support global location (fixes #502) (ff906fb)
- add retry logic for missing X-Goog-Upload-Status header for js (8cf039e)
0.12.0 (2025-04-30)
- add support for live grounding metadata (f5ed429)
0.11.0 (2025-04-30)
- add models.delete and models.update to manage tuned models (0766718)
- Added support for Tuning operations (d5a035f)
- make min_property, max_property, min_length, max_length, example, patter fields available for Schema class when calling Gemini API (5f91ee7)
- Apply converters to list items when the API value isn't an array (249769f)
- chats: Properly handle streaming errors to not throw an error that couldn't be caught, which might result in crash in web. Fixes #487 (0b62e15)
- CI: Fix docs generation for release event (899969e)
- Clean the CHANGELOG to remove the changes not included in the bundle. (89b1d66)
- do not raise error for
defaultfield in Schema for Gemini API calls (6f72396) - Don't transform lists twice (249769f)
0.10.0 (2025-04-23)
- add additional realtime input fields (2150416)
- Add helper function
GenerateContentResponse.datato return concatenation of all inline data parts. (97b59a2) - Allow users to set AbortSignal inside per request config to cancel http request (c44f48d)
- support
defaultfield in Schema when users call Gemini API (2e61cce) - Support setting the default base URL in clients via setDefaultBaseUrls() (df978b8)
- Use ws:// for websockets established over http base URLs. (774505b)
- Return actual error text from streaming response (808e0b5), closes #346
- Update _api_client to parse and throw errors during processing stream responses (fixes #461) (1932f1d)
0.9.0 (2025-04-17)
- Simplified the types allowed on the generateContent contents parameter.
- add a helper module to process zod objecsts. (fad2789)
- add support for model_selection_config to GenerateContentConfig (37a9cae)
- Expose transcription configurations for audio in TS, and move generationConfig to the top level LiveConnectConfig (d3a841d)
- Simplified the types allowed on the generateContent contents parameter. (324d158)
- Support audio transcription in Vertex Live API (8d11c04)
- Support RealtimeInputConfig, and language_code in SpeechConfig in python (004ff6a)
- Update VertexRagStore (043d06c)
- CI: Fix stable docs generation with release event (89c61b9)
0.8.0 (2025-04-09)
- Add domain to Web GroundingChunk (dc56670)
- Add generationComplete notification to LiveServerContent (4d9811b)
- add session resumption to live module (b5c6975)
- add session resumption. (fff2b66)
- Add thinking_budget to ThinkingConfig for Gemini Thinking Models (76e0e18)
- Add traffic type to GenerateContentResponseUsageMetadata (d84156e)
- Add types for configurable speech detection (fc861bc)
- Add types to support continuous sessions with a sliding window (0fd8256)
- Add UsageMetadata to LiveServerMessage (67b34f7)
- expose generation_complete, input/output_transcription & input/output_audio_transcription to SDK for Vertex Live API (1e8be50)
- merge GenerationConfig into LiveConnectConfig (d25d77d)
- Populate X-Server-Timeout header when a request timeout is set. (6f00495)
- support media resolution (9ebd58b)
- Support models.get() for getting model information (fc62381)
- Update Live converters to pass along configurable speech detection parameters (8301fa2)
- Update Live converters to pass along params to support continuous sessions with a sliding window (3814d92)
- Use authentication headers as provided by google-auth-library (94b11a1)
0.7.0 (2025-03-27)
- Change File.sizeBytes from number type to string type
- Add experimental generate_video support (0fa1f05)
- add MediaModalities for ModalityTokenCount (9869098)
- Change File.sizeBytes from number type to string type (184c7db)
- Update tLiveClienttToolResponse() to accept FunctionResponse[] (4cab8bf)
0.6.1 (2025-03-25)
- Add engine to VertexAISearch (69dfbaf)
- allow title property to be sent to Gemini API. Gemini API now supports the title property, so it's ok to pass this onto both Vertex and Gemini API. (c5855a3)
- implement files.delete for JS client SDK. (4ac44de)
- Save prompt safety attributes in dedicated field for generate_images (1a774fa)
- schema transformer logic fix. (6311f60)
0.6.0 (2025-03-20)
- Unexport Content converter functions
- add IMAGE_SAFTY enum value to FinishReason (81ae907)
- Separate converter functions to dedicated _{module}_converters.ts file for readability (bb9ba98)
0.5.0 (2025-03-20)
- Make "turnComplete:true" the default.
- Add sendClientContent, sendRealtimeInput, sendToolResponse to live session (e7ec3c0)
- Make "turnComplete:true" the default. (5f77e3e)
- Support Google Cloud Express for Vertex AI (e15c7f3)
- support new UsageMetadata fields (fe000ed)
- Support Vertex AI on browser runtimes (e15c7f3)
- Upgrade the SDK launch stage to preview. (da38b6d)
0.4.0 (2025-03-14)
- remove the createPartFromVideoMetadata usability function.
- enable union type for Schema when calling Gemini API. (180983c)
- Provide a better error message when trying to use VertexAI in browsers. (1ab1402)
- Support returned safety attributes for generate_images (a0e0fcf)
- throw exception when given method is not supported in Gemini API or Vertex AI (96ccb6f)
- remove the createPartFromVideoMetadata usability function. (d660a7f)
- Make file.upload use named parameters.
- Enable Live for Vertex AI. (2bda9d4)
- Set web as the browser entry points for bundlers that don't support the exports key (18cb728)
- Make file.upload use named parameters. (60433f4)