Changelog
2.0.1 (2026-05-09)
Bug Fixes
- Update response_format field names to snake_case. (97142b1)
2.0.0 (2026-05-07)
⚠ BREAKING CHANGES - Interactions Only
Note: The breaking changes are only in interactions. GenerateContent usage in unaffected.
Refer to https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026 for details
- Add steps for interactions (41d8e5b)
- Rename SSE events to interaction.created and interaction.completed (0549c1c)
- Deprecate legacy response_format and publish new polymorphic field. (7fdc9a1)
Documentation
1.75.0 (2026-05-04)
Features
- [Python] Multimodal file search (fe5d69e)
Bug Fixes
- Avoid caching stale token in async mTLS path (4e17a9c)
Documentation
- Fix python docstrings for Image.from_file() to use kwargs (cce5398)
- Update docs (ef26583)
1.74.0 (2026-04-29)
Features
- [Interactions] Add FileCitation.{custom_metadata,media_id,page_number} (aed41ec)
- Add
output_info to BatchJob (7b77ab8)
- Add gemini-3.1-flash-tts-preview model to options (8bdc1c3)
- Add ImageResizeMode for GenerateVideos (317d2af)
- Add new Gemini Deep Research agent models (16fffbd)
- Add one_of support to JSONSchema for Agent Platform (8c00c52)
- Add Vertex Dataset input and output options for batch jobs (d880f92)
- interaction-api: Add grounding tool usage breakdown to Interaction Usage. (b24fb5a)
- Introduce
enterprise to Client constructor and GOOGLE_GENAI_USE_ENTERPRISE (693fd9a)
- Replace the more ambiguous rate field with sample_rate. (88d9b4a)
Bug Fixes
- Catch google-auth wrapped errors (48ac850)
- Removing Python 3.9 support due to EOL (8bc2b10)
- retry: Retry on httpx.TimeoutException with HttpRetryOptions (#2345) (0598bab)
- Streaming method doesn't handle multi-line SSE (f8a2e7e)
- Typing in
AsyncClient.__aexit__, __exit__. (a74dc65)
Documentation
- Add instruction for custom endpoint (dd79904)
- Fix broken link for rate limits (d22ea99)
- Regenerate docs for 1.73.1 (2fb714b)
- Remove duplicate line in genai client docstring. (b1c6026)
- Replace Vertex AI with Gemini Enterprise Agent Platform (7c1ecd5)
- Update doc string to replace
Vertex AI with Gemini Enterprise Agent Platform, update method error message to replace Vertex AI with Gemini Enterprise Agent Platform (previously known as Vertex AI), update converter error message to replace Vertex AI with Gemini Enterprise Agent Platform (413f0f9)
- Update Gemini Enterprise Agent Platform home page url (b02cb95)
- Update README.md with correct Pydantic link (fa97cc6)
- Update README.md with correct Pydantic link (fa97cc6)
1.73.1 (2026-04-14)
Bug Fixes
- Refactor Webhook types in GenAI SDKs for easier useage (3f36ca1)
- Rename
webhooks.retrieve to webhooks.get. (649f4b0)
Documentation
- Update python docs for 1.73.0 (acd3767)
1.73.0 (2026-04-13)
Features
- Add DeepResearchAgentConfig fields (ec8ca87)
- Add webhook and webhookConfig for js and python sdk (ccec350)
- Add webhook_config to batches.create() and models.generate_videos() (772d2fc)
- Wire the webhook into python and js client. (841bf22)
Bug Fixes
- Refine Pyink blank line insertion logic and fix range-based formatting regressions. (b91bda5)
1.72.0 (2026-04-09)
Features
- Add "eu" as a supported service location for Vertex AI platform. (888a731)
- Add Live Avatar new fields (ad1777e)
- Add support for new audio MIME types: opus, alaw, and mulaw (74eb373)
- Add the delete method for the Agent Engine Task Store Service (d821082)
Documentation
- Update python docs for 1.71.0 (e82f9fc)
1.71.0 (2026-04-08)
Features
- Introduce TYPE_L16 audio content and optional fields. (07e932f)
Documentation
- Remove deprecated product recontext model samples from docstrings (aca7dcf)
1.70.0 (2026-03-31)
Features
- Support dedicated TextAnnotationDelta for streaming tool responses (5c820f2)
Bug Fixes
1.69.0 (2026-03-27)
Features
- Add consent_audio and voice_consent_signature and AsyncSession.setup_complete (69a02c4)
- Add custom_metadata to FileSearchResult. (aed1559)
- Add labels field to Veo configs (208a173)
- Add mime type for Audio content (674b837)
- Add model_status to GenerateContentResponse (Gemini API only) (ce86f2b)
- Add part_metadata in Part (Gemini API only) (ce86f2b)
- Add service tier for interactions. (b07002e)
- Add service tier to GenerateContent. (12b404b)
- Add support for more image and audio MIME types in Interactions content (8ec977c)
- Add supported models to the ModelOptions (1ccad7b)
- Autoenable mTLS in environment with bound token (Agent Engine with AgentAuthority) through google-auth migration (except custom client args, custom client or custom ClientSession) (a95d08a)
- genai: Add TURN_INCLUDES_AUDIO_ACTIVITY_AND_ALL_VIDEO to TurnCoverage (1ba8e2e)
- Support hyperparameters in distillation tuning (ad38e3a)
Bug Fixes
- Support us region routing (8e3e00c)
Documentation
- Update python docs for 1.68. (07ae1b1)
1.68.0 (2026-03-17)
Breaking changes
- [Interactions] Breaking change to Interactions API to refactor TextContent annotations to use specific citation types (6c3379f)
- [Interactions] Breaking change for Interactions, rename ContentDelta unions. (1b03909)
- [Interactions] Breaking change to Interactions API to rename rendered_content to search_suggestions (0e21c4e)
Features
- [Interactions] Add and update 'signature' fields for tool call/result content types. (d896373)
- [Interactions] Support Google Maps in Interactions (68f247c)
- Support include_server_side_tool_invocations for genai. (546440c)
Bug Fixes
Documentation
- Regenerate docs for 1.67.0 (ff7469a)
1.67.0 (2026-03-12)
Features
- Add inference_generation_config to EvaluationConfig for Tuning (1fdb4b8)
- Add live history_config with initial_history_in_client_content (a80babd)
- Add support for referencing registered metrics by resource name in evaluation run API (41b348e)
- Enable language code for audio transcription config in Live API for Vertex AI (c04be0d)
Bug Fixes
- Forward http_options in async_request_streamed to enable retry support (8b3be87)
- Forward http_options in async_request_streamed to enable retry support (#2097) (8b10efb)
1.66.0 (2026-03-03)
Features
- Add gemini-3.1-flash-image-preview model (dd52cc2)
- Support signature for all Interaction tool types (abb388e)
- Update data types from discovery doc. (15666c0)
1.65.0 (2026-02-26)
Features
- Add gemini-3.1-pro-preview to list of models in Interactions (fe86870)
- Add Image Grounding support to GoogleSearch tool (0035182)
- Enable server side MCP and disable all other AFC when server side MCP is configured. (4dd7b16)
- Support more image sizes and resolutions (8b2a4e0)
Bug Fixes
- Change interactions media mime type to enum (breaking change for experimental feature) (e0f3378)
- Handle non-list response_stream in HttpResponse.json property (006042d)
- Handle non-list response_stream in HttpResponse.json property (#1903) (61aec34)
- Make aiohttp an optional dependency, bump aiohttp version upperbound, and ensure HttpOptions.async_client_args propagates to custom aiohttp.ClientSession request args for proxy etc. (b28d144), closes #2090 #2051 #1950
Documentation
- Regenerate docs for 1.64.0 (a023141)
- Update README and codegen_instructions for structured outputs (4d5a978)
1.64.0 (2026-02-18)
Features
- Add UnifiedMetric support to Vertex Tuning evaluation config (9a9908a)
- Support multimodal embedding for Gemini Embedding 2.0 and support MaaS models in Models.embed_content() (Vertex AI API) (af40cc6)
1.63.0 (2026-02-11)
Features
- Add INCOMPLETE status to Interaction. (1a84605)
- Support encryption_spec in tuning job creation configuration for GenAI SDK (057d6f0)
Bug Fixes
- Base_url and global location parsing (2c40555)
- Remove build warning due to extra comma: assertion on a tuple is always true (1cc2c2d)
- Remove debug print statement. (60c0a2f)
- Remove unused import of
websockets (3b82b5f)
Documentation
- Add docstrings to C# and Python SDK Operations methods. (f2e85a3)
- Regenerate docs for 1.62.0 (455ec26)
1.62.0 (2026-02-04)
Features
- Update data types from discovery doc. (089ba97)
Bug Fixes
- Add error handling for live and live music APIs (1148276), closes #668
1.61.0 (2026-01-30)
Features
- Add
include_input query parameter to Get Interaction endpoint. (a0240d9)
- Add registerFiles for you can use gcs files with mldev. (965395b)
- Support distillation tuning (9e49d71)
- Support OSS Tuning in GenAI SDK (51748a7)
Bug Fixes
- Add metadata in batch inlined response (08c47aa)
1.60.0 (2026-01-21)
Features
- Add ModelArmorConfig support for prompt and response sanitization via the Model Armor service (8d1091a)
- Update data types from discovery doc. (4209289)
- Update data types from discovery doc. (7db2c2d)
Documentation
- Regenerate docs for 1.59.0 (351e490)
- Update docs to include interactions and file_search_stores module (a21841c)
1.59.0 (2026-01-15)
Features
- Set the environment variable GOOGLE_API_PREVENT_AGENT_TOKEN_SHARING_FOR_GCP_SERVICES to 'false' within BaseApiClient to disable bound token sharing. (79ac880)
- Support 4:5 and 5:4 aspect ratio in Interactions (1ddd9f1)
Documentation
- Regenerate docs for 1.58.0 (39a8b06)
1.58.0 (2026-01-14)
Features
- Add FileSearchCallContent to Interactions (a882dea)
- Add ImageConfig to GenerationConfig for image generation in Interactions (b61163f)
- Support passing the custom aiohttp.ClientSession through HttpOptions.aiohttp_client (750648f), closes #1662
- Voice activity support (b7b1c2e)
Bug Fixes
- Serialize Pillow images losslessly by default (8d7c74d)
Documentation
- Regenerate docs for 1.57.0 (65018b6)
1.57.0 (2026-01-07)
Features
- [Python] add RegisterFiles so gcs files can be used with genai. (68fa075)
- Add gemini-3-pro-preview support for local tokenizer (48f8256)
- Add PersonGeneration to ImageConfig for Vertex Gempix (c66e0ce)
Bug Fixes
- Remove validation for empty text parts on Chat, this will support keeping the history in chat when the API yields back such a part. (215c852)
Documentation
- Regenerate docs for 1.56.0 (b4c063e)
- Update
codegen_instructions.md for Gemini 3 Flash (22500b5)
- Update Virtual Try-On model id in samples and docstrings (5bf4d62)
1.56.0 (2025-12-16)
Features
- Add minimal and medium thinking levels. (96d644c)
- Add support for Struct in ToolResult Content. (8fd4886)
- Add ultra high resolution to the media resolution in Parts. (356c320)
- Add ULTRA_HIGH MediaResolution and new ThinkingLevel enums (336b823)
- Define and use DocumentMimeType for DocumentContent (dc7f00f)
- Support multi speaker for Vertex AI (ecb00c2)
Bug Fixes
- Api version handling for interactions. (436ca2e)
Documentation
- Add documentation for the new Interactions API (Preview). (e28a69c)
- Update and restructure codegen_instructions (00422de)
- Update docs for 1.55 (1cc43e7)
1.55.0 (2025-12-11)
Features
- Add the Interactions API (836a3)
- Add enableEnhancedCivicAnswers feature in GenerateContentConfig (15d1ea9)
- Add IMAGE_RECITATION and IMAGE_OTHER enum values to FinishReason (8bb4b9a)
- Add voice activity detection signal. (feae46d)
Bug Fixes
- Replicated voice config bytes handling (c9f8668)
Documentation
- Regenerate docs for 1.54.0 (8bac8d2)
1.54.0 (2025-12-08)
Features
- Support ReplicatedVoiceConfig (07c74dd)
Bug Fixes
- Apply timeout to the total request duration in aiohttp (a4f4205)
- Make APIError class picklable (fixes #1144) (e3d5712)
Documentation
- Regenerate docs for 1.53.0 (3a2b970)
1.53.0 (2025-12-03)
Features
- Add empty response for tunings.cancel() (97cc7e4)
Bug Fixes
- Convert 'citationSources' key in CitationMetadata to 'citations' when present (fixes #1222) (2f28b02)
- Fix google.auth.transport.requests import error in Live API (a842721)
Documentation
- Improve docs for google.genai.types (5b50adc)
- Recommend using response_json_schema in error messages and docstrings. (c0b175a)
- Updating codegen instructions to use gemini 3 pro and nano banana pro (060f015)
1.52.0 (2025-11-21)
Features
- Add support for configuring resource scope when using base_url (a3e0859)
Bug Fixes
TypeError: issubclass() arg 1 must be a class when usingList[str] for contents (c624d7e)- Create new aiohttp Client Session if loop is closed (1dc35ea)
1.51.0 (2025-11-18)
Features
- Add a pre-validation hook to warn about Pydantic model type mismatches. (f7af6ef)
- Add display name to FunctionResponseBlob (52906d5)
- Add display name to FunctionResponseFileData (7c39f70)
- Add generate_content_config.thinking_level (30b00db)
- Add image output options to ImageConfig for Vertex (014aaad)
- Add part.media_resolution (30b00db)
- Support Function call argument streaming for all languages (9b2ca50)
Bug Fixes
- Only log warnings once for accessors in GenerateContentResponse and LiveServerMessage (eec841e)
Documentation
- Remove gemini 2 below model reference (c42ddff)
- Update sample code to use gemini 2 or above (d48bbba)
- Update Veo and Imagen model id in README (5c69122)
1.50.1 (2025-11-13)
Bug Fixes
- Do not use ADC if passing a base_url, no project, no location (a00b67a)
- Ensure the custom httpx client and async client won't be closed automatically (9a9fa3c)
1.50.0 (2025-11-12)
Features
- Use pytest-xdist for test parallelization (6ff82fc)
Bug Fixes
- Add missing fields to the model types (4b855e6)
- Don't generate warnings from response.text property because of thought_signature. (dd9360d)
- Fix base_steps parameter for recontext_image (d94077b)
- Fix file_search_stores.documents pagination (8d40d48)
- Fix models.list() filter parameter (7fa1e41)
- Handle SSE error message types properly in streaming (72afa50)
- Roll back a breaking change to the import system (8674003)
Documentation
- Log Schema.json_schema and Schema.from_json_schema only once (517300c)
- Regenerate docs for 1.49.0 (74b680a)
1.49.0 (2025-11-05)
Features
- Add complete stats to BatchJob (b211466)
- Add FileSearch tool and associated FileSearchStore management APIs (2e2a78d)
- Add FileSearch tool and associated FileSearchStore management APIs (7370d24)
- Add image_size to ImageConfig (Early Access Program) (81c027c)
- Make genai.Part constructible from PartUnionDict. (7526e4d)
Bug Fixes
- Raise errors during file upload. (946a17e)
- Use duck type in t_part transformer (be82981)
Documentation
- Log deprecation waring to Schema.from_json_schema method and Schema.json_schema method (856789a)
- Regenerate docs for 1.48.0 (004d238)
- Update log message in Schema.json_schema and Schema.from_json_schema methods (5bf69e5)
1.48.0 (2025-11-03)
Features
- Added phish filtering feature. (a9297b7)
- Drop support for Python 3.9 - EOL (b542082)
Bug Fixes
- Append the current model chunk to contents in async streaming (7c5cf56)
- Disable AFC when there are incompatible tool presented. (ce13aef)
- Offload sync python tool calls to a thread when used from async context. (f2a0782)
- Only show warning log for non-text or multi-candidate fields when response.text is accessed directly (bf82505)
1.47.0 (2025-10-29)
Features
- Add safety_filter_level and person_generation for Imagen upscaling (6196b1b)
- Add support for preference optimization tuning in the SDK. (4540f9d)
- Pass file name to the backend when uploading with a file path (4fa2edd)
- Support default global location when not using api key with vertexai backend (6340ce0)
- Support retries in API requests (ac70ecd)
Bug Fixes
- Check environment Vertex AI api key for credential precedence (9bd758c)
- Correct pydantic version range (bytes fields are broken with pydantic<=2.8). (d24cb56)
- Make
__del__ methods more robust in _api_client and client. (64cab58)
- Setting custom httpx async client will ensure using httpx client even if aiohttp is installed (7bd1bde)
- Support custom_base_url for Live and other APIs when project/location are unset and even when project/location are set in the environment, and avoid Automatic Default Credentials (7bd1bde)