NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI · #151 most downloaded on PyPI
GenAI Python SDK
Last release 8 days ago
22 Sep 2026
Ships on a steady schedule
a new release about every 2 weeks
Nearly every release is documented
notes for 60 of the last 60 stable releases
3 versions withdrawn
withdrawn after publishing
2 years old
125 releases · first in 2024
One column per month.
Drop support for Python 3.9 - EOL
Add safety_filter_level and person_generation for Imagen upscaling
__del__ methods more robust in _api_client and client. (64cab58)Add enable_enhanced_civic_answers in GenerationConfig
Keys in Live API tool responses are incorrectly re-cased
Support fully override base_url and raw model name when none of the project, locations, api_key are configured
Enable Google Maps tool for Genai.
Add labels field to Imagen configs
FunctionResponsePart and creating FunctionResponse Part with FunctionResponseParts (72c92d8)READ_BUFFER_SIZE in _api_client.py for streaming large chunks in new model (981bba7)Add NO_IMAGE enum value to FinishReason
Add ImageConfig to GenerateContentConfig
Unbreak client closed errors when using vertexai session service
Add FunctionResponsePart & ToolComputerUse.excludedPredefinedFunctions
JOB_STATE_RUNNING and JOB_STATE_EXPIRED for Gemini Batches states (739f72d)pre_tuned_model_checkpoint_id from tuning config. (1d3d28a)Add 'turn_complete_reason' and 'waiting_for_input' fields.
Add VideoGenerationMaskMode enum for Veo 2 Editing
VideoGenerationMaskMode enum for Veo 2 Editing (3d73cc5)Add labels to create tuning job config
Support Veo 2 Editing on Vertex
[Python] Implement async embedding batches for MLDev.
Add resolution field for Gemini Developer API Veo 3 generation
Add sdkHttpResponse.headers to *Delete responses.
sdkHttpResponse.headers to *Delete responses. (0101d47)Support Imagen image segmentation on Vertex
Add evaluation support to Vertex tuning
Add image_size field for Gemini Developer API Imagen 4 generation
Add images quick accessor to GenerateImagesResponse
Add image_size field for Vertex Imagen 4 generation
Add addWatermark parameter to the edit image configuration.
addWatermark parameter to the edit image configuration. (a5b1545)Add new languages for Imagen 4 prompt language
types more self-contained (7d0a7d8)Support different media input in Vertex Live API
Enable Vertex Multimodal Dataset as input to supervised fine-tuning.
Add compressionQuality enum for generate_videos
Re-raise exception during retries
Enable json schema for controlled output and function declaration.
Add datastore_spec field for VertexAISearch
Add enhance_prompt field for Gemini Developer API generate_videos
GEMINI_API_KEY as environment variable for setting API key. (ae2392c)Adding thought_signature field to the Part to store the signature for thoughts.
thought_signature field to the Part to store the signature for thoughts. (303f906)Remove unsupported Lyria enum for music generation mode
Fix broken tool use in generate content.
Add time range filter to Google Search Tool
time range filter to Google Search Tool (b79c414)Support display_name for Blob class when calling Vertex AI
Add Tool.enterprise_web_search field
Tool.enterprise_web_search field (731c5a3)Add models.delete and models.update to manage tuned models
default field in Schema for Gemini API calls (1d3d1c9)propertyOrdering when schema is specified as dict or types.Schema. (48eebe0)## 1.12.1 (2025-04-24) ### Bug Fixes * Fix required import of Pillow (f23772d) * Remove __del__ in client
Add additional realtime input fields
default field in Schema when users call Gemini API (1e56add)Add support for model_selection_config to GenerateContentConfig
remove Part.from_video_metadata
Add specialized send methods to the live api
send methods to the live api (9c4e4dc)Log warning to users that Part.from_video_metadata will be deprecated.
types.Content is passed to send_message() to correctly adhere to type annotation. To migrate code previously passing content = types.Content(...) to send_message(), pass content.parts instead.(7fb2c7e)Leverage httpx connection pooling and avoid instantiation of httpx.Client or httpx.AsyncClient for each call
types.Content is passed to send_message() to correctly adhere to type annotation. To migrate code previously passing content = types.Content(...) to send_message(), pass content.parts instead.Add base steps to EditImageConfig
Determine mime_type from local images
Add response_id and create_time to GenerateContentResponse
Add generate_videos (Veo 2) support for Python
Enable Media resolution for Gemini API.
Add support for typing.Literal in response_schema (fixes #264)
Tuning.tune for Gemini API no longer blocks until the tuning operation is resolved (fcf8888)Remove deprecated field: deprecated_response_payload
deprecated_response_payloadANY mode. (451cf98)deprecated_response_payload (197fa46)Add labels for GenerateContent requests
Rename files.upload argument to "file" instead of "path".
Moved the HttpOptions class into types.py, by making it autogenerated. This causes the following breaking change:
Callable type not just functions.Callable type not just functions. (0e4a003)Renamed types.FunctionDeclaration.from_function to types.FunctionDeclaration.from_callable
types.FunctionDeclaration.from_function to types.FunctionDeclaration.from_callable (5afa6bb)bytes type handling (the base64.urlsafe_bencode error) (1bc161d)Support API keys for VertexAI mode generate_content
ser_json_byte val_json_bytes in bytes type public interface (1176e43)Make Imagen upscale_factor a required argument, make upscale config optional
contents must be passed to CreateCachedContentConfig instead of as a parameter to create_cached_content.
Your coding agent can read these notes before it upgrades. Set up the MCP server →