NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI · #288 most downloaded on PyPI
Vertex AI API client library
Last release 4 days ago
30 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
2 versions withdrawn
withdrawn after publishing
6 years old
223 releases · first in 2020
One column per quarter.
Add serving profiles support to agentplatform genai sdk
Allow unknown union variants when validating gaos responses in evals
Add psc_endpoint parameter to Sandboxes.send_command and Sandboxes.generate_browser_ws_headers so VPC-SC sandboxes with ingressControlConfig.enablePri
psc_endpoint parameter to Sandboxes.send_command and Sandboxes.generate_browser_ws_headers so VPC-SC sandboxes with ingressControlConfig.enablePrivateServiceConnect=true (which populate only connection_info.service_attachment, never load_balancer_hostname / load_balancer_ip) can route data-plane traffic through a customer-provisioned PSC endpoint. Existing non-VPC-SC callers are unaffected; the parameter defaults to None and the load-balancer branches remain the primary path. (39971ae)_gcs_utils module. (7c94277), refs #5928 #7129Add execute_bash to the Agent Engine sandbox SDK.
ingress_control_config when creating a SandboxEnvironmentTemplate via the SDK (Python, Java, JS). (2dccde8)Note: This release contains all updates originally intended for v2.0.0. The version was bumped to 2.0.1 due to a PyPI registry conflict with a previou
Note: This release contains all updates originally intended for v2.0.0. The version was bumped to 2.0.1 due to a PyPI registry conflict with a previously yanked 2.0.0 package.
agent_engines module renamed to runtimes.a2a tasks module is removed.sessions and sandboxes moved to top-level modules.runtime templates import changed from agentplatform.agent_engines.templates to agentplatform.frameworks.evals.run_inference(agent=...) now accepts types.Runtime instead of types.AgentEngine.client.memory_banks sub-module.google-cloud-agentplatform. (e0e32c2)agentplatform: agent_engines module renamed to runtimes .
agent_engines module renamed to runtimes.a2a tasks module is removed.sessions and sandboxes moved to top-level modules.runtime templates import changed from agentplatform.agent_engines.templates to agentplatform.frameworks.evals.run_inference(agent=...) now accepts types.Runtime instead of types.AgentEngine.client.memory_banks sub-module.google-cloud-agentplatform. (e0e32c2)chore(main): release 2.0.0
Update CHANGELOG.md
Co-authored-by: release-please[bot] <55107282+release-please[bot]@users.noreply.github.com>
Co-authored-by: Ayush Agrawal 108827967+speedstorm1@users.noreply.github.com
Release was yanked due to a PyPI registry conflict with a previously pulled version. All changes intended for 2.0.0 are included in the 2.0.1 release above.
Truncate default pipeline and staging bucket names to 63 chars
Add enable_data_retention to ToolParallelAiSearch, Add step_count to ReinforcementTuningHyperParameters, Add BidiGenerateContentSetup
Add build configuration support to Vertex AI Reasoning Engine
iam_credentials_v1 lazily in the sandboxes module. (2a18bff)google-cloud-iam dependency from the agent_engines extra. (25deb44)order_by in message .google.chat.v1.SearchSpacesRequest is changed (19144a4)query in message .google.chat.v1.SearchSpacesRequest is changed (19144a4)spaces in message .google.chat.v1.SearchSpacesResponse is changed (19144a4)use_admin_access in message .google.chat.v1.SearchSpacesRequest is changed (19144a4)SearchSpaces in service ChatService is changed (19144a4)GenAI Client(evals) - add create EvaluationExperiment
autogen distribution instead of ag2 for the AG2 integration. (dbcf815)Add detailed agent tool definitions for display
Remove OpenTelemetry event logger setup since the Events API/SDK has been deprecated and is soon being removed entirely.
Add revision_labels, revision_expire_time, revision_ttl, disable_memory_revisions, metadata, metadata_merge_strategy and overlap_event_count to Memory
Add deprecation warning to vertexai.rag module in favor of agentplatform.Client().rag
datasets.assemble now returns tuple[str, Optional[bigframes.pandas.DataFrame]] (the BigQuery table id without the bq:// prefix, and the assembled table as a BigFrames DataFrame) instead of the BigQuery URI str. By default no DataFrame is loaded (load_dataframe=False): the second element is None, no BigQuery read is performed, and bigframes is not required. Pass load_dataframe=True to load the assembled table.datasets.assemble. (8e52bd2)Add ReasoningEngineRuntimeRevisionService and update related fields in ReasoningEngine
GenAI client - Add delete_file and delete_corpus methods to RAG module
to_batch_job_source and get_batch_job_destination to MultimodalDataset (116b71c)create_from_gemini_request_jsonl in Datasets and AsyncDatasets classes. (6e51cdf)GenAI SDK client - add rag.create_corpus() method
Add Skill Registry GetSkillRevision and ListSkillRevisions to SDK
http_options from AgentEngine.query family of methods. (e6818d3)Add Skill Registry module to Vertex AI client - Get Skill method
skills.create (bc4be37)Remove google-genai version contraints
GenAI Client(evals) - Add red_teaming_config support for create_evaluation_run
Add GetSkillRevision and ListSkillRevisions methods in Vertex AI Skill Registry SDK
Add client for Vertex AI Skill Registry - Get Skill method
reference column visualization for eval_result (42de11b)Add Agent Engine Runtime Versioning support to SDK.
Unpin litellm upper bound to allow CVE-2026-35030 remediation
create_* in some cases (ca6b45e)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 (18d97b6)Disable bound token sharing check for agent_engines
UpdateMultimodalDatasetConfig and replace with VertexBaseConfig. (f5dc932)Methods for multimodal datasets now require the full resource name instead of just the dataset ID.
datasets module. (505aa8c)The labels field has been removed from GeminiExample. The tools and safety_settings fields have been turned into lists.
to_bigframes has been removed from the datasets module and moved into the MultimodalDataset class. Instead of dataframe = client.datasets.to_bigframes
to_bigframes has been removed from the datasets module and moved into the MultimodalDataset class. Instead of dataframe = client.datasets.to_bigframes(multimodal_dataset=multimodal_dataset), use dataframe = multimodal_dataset.to_bigframes() to create a BigFrame instance from a multimodal dataset.Add session_id to Create Session to allow custom session id
Breaking change in experimental methods: create_from_bigquery and update_multimodal_dataset no longer automatically prepend a missing bq:// prefix for…
create_from_bigquery and update_multimodal_dataset no longer automatically prepend a missing bq:// prefix for BigQuery URIs. When using the new function MultimodalDataset.set_bigquery_uri the prefix will still be added if needed.metadata_filter usage to all retrieval and generator methods in rag_retrieval.py (841c597)MultimodalDataset. (e164b19)create_from_pandas with async versions. (2767273)Add AgentEngine Sessions module
gemini_request_read_config instead of template_config in all functions. (f138162)GenAI SDK client - Breaking Change: Remove the following types from vertexai.types and depend on these types from the Google GenAI SDK via google.gena…
retrieve_contexts_async and ask_contexts SDK methods in rag_retrieval.py (0e0137e)VALIDATED Function Calling mode to v1 (981a551)embed_content_config to message EmbedContentRequest (981a551)embed_content_config to message EmbedContentRequest (981a551)agent parameter to create_evaluation_run methods, rather than being part of the AgentInfo object. This parameter is now required if agent_info is provided (dab185a)rag_file_metadata_config across ImportRagFilesRequest and UploadRagFileRequest (981a551)"global" endpoint supports "grpc" transport
context and history columns when creating Evaluation run from dataframe (e887a2e)agent_data in EvaluationDataset in create_evaluation_run (2b0a98c)A new field computation_based_metric_spec is added to message .google.cloud.aiplatform.v1beta1.Metric
computation_based_metric_spec is added to message .google.cloud.aiplatform.v1beta1.Metric (e5f71de)evaluation_run is added to message .google.cloud.aiplatform.v1beta1.EvaluateDatasetRun (e5f71de)inference_generation_config is added to message .google.cloud.aiplatform.v1beta1.EvaluationConfig (e5f71de)llm_based_metric_spec is added to message .google.cloud.aiplatform.v1beta1.Metric (e5f71de)metric_results is added to message .google.cloud.aiplatform.v1beta1.EvaluateInstancesResponse (e5f71de)predefined_metric_spec is added to message .google.cloud.aiplatform.v1beta1.Metric (e5f71de)ComputationBasedMetricSpec is added (e5f71de)LLMBasedMetricSpec is added (e5f71de)MetricResult is added (e5f71de)PredefinedMetricSpec is added (e5f71de)traffic_type, tool_use_prompt_tokens_details to message GenerateContentResponse.UsageMetadata (e5f71de)name in message .google.cloud.aiplatform.v1beta1.Dataset (e5f71de)name in message .google.cloud.aiplatform.v1beta1.DatasetVersion (e5f71de)google.cloud.aiplatform.v1beta1.BatchPredictionJob (e5f71de)google.cloud.aiplatform.v1beta1.BatchPredictionJob (e5f71de)MULTI_TURN_TOOL_USE_QUALITY, MULTI_TURN_TRAJECTORY_QUALITY, MULTI_TURN_TASK_SUCCESS). (513d4d6)AggregationMetric is changed (e5f71de)aggregation_output in message .google.cloud.aiplatform.v1beta1.EvaluateDatasetResponse is changed (e5f71de)expire_time in message .google.cloud.aiplatform.v1beta1.Session is changed (e5f71de)filter in message .google.cloud.aiplatform.v1beta1.ListSessionsRequest is changed (e5f71de)output_info in message .google.cloud.aiplatform.v1beta1.EvaluateDatasetResponse is changed (e5f71de)ttl in message .google.cloud.aiplatform.v1beta1.Session is changed (e5f71de)EvaluateDatasetResponse is changed (e5f71de)Metric is changed (e5f71de)OutputInfo is changed (e5f71de)evaluate_dataset_response clarified (e5f71de).google.cloud.aiplatform.v1.DatasetService, .google.cloud.aiplatform.v1.Dataset, and .google.cloud.aiplatform.v1.DatasetVersion (e5f71de).google.cloud.aiplatform.v1beta1.Dataset and .google.cloud.aiplatform.v1beta1.DatasetVersion (e5f71de)Add max_concurrent_active_run_count support to Scheduled Pipelines client GA.
Add support for BYO-dockerfile in AE deployment
Deprecate prompt_optimizer.optimize and prompt_optimizer.optimize_prompt in favor of prompts.launch_optimization_job and prompts.optimize
Add fps to message VideoMetadata
Add ComputationBasedMetricSpec to support metrics like EXACT_MATCH, BLEU, and ROUGE in EvaluationRun configurations.
ComputationBasedMetricSpec to support metrics like EXACT_MATCH, BLEU, and ROUGE in EvaluationRun configurations. (9d32dd5)GenAI Client(evals) - Add support for inference_configs in create_evaluation_run.
inference_configs in create_evaluation_run. (33fe72a)Deprecate tuning public preview SDK in favor of tuning SDK
optimize_prompt method (715cc5b)examples_dataframe type to PandasDataFrame in Prompt Optimizer. (a2564cc)Add Lustre support to the Vertex Training Custom Job API
GenAI Client(evals) - patch for vulnerability in visualization
A new field min_gpu_driver_version is added to message .google.cloud.aiplatform.v1beta1.MachineSpec
min_gpu_driver_version is added to message .google.cloud.aiplatform.v1beta1.MachineSpec (26dfdfe)An existing field transfer_to_agent is removed from message .google.cloud.aiplatform.v1beta1.EventActions
transfer_to_agent is removed from message .google.cloud.aiplatform.v1beta1.EventActionsbigtable_metadata field name in FeatureOnlineStoreenableDirectBigtableAccess field name in FeatureOnlineStore`bigtable_metadata field name in FeatureViewgpu_partition_size in machine_spec v1 api (e0bc3d8)ReplicatedVoiceConfig to VoiceConfig to enable Gemini TTS voice replication (e0bc3d8)ReplicatedVoiceConfig to VoiceConfig to enable Gemini TTS voice replication (e0bc3d8)SUCCESSFULLY_DEPLOYED and FAILED_TO_DEPLOY to DeploymentStage (e0bc3d8)SUCCESSFULLY_DEPLOYED and FAILED_TO_DEPLOY to DeploymentStage (e0bc3d8)transfer_to_agent is removed from message .google.cloud.aiplatform.v1beta1.EventActions (e0bc3d8)adk deploy agent_engine (9301551)from vertexai.types import TypeName without needing to run from vertexai import types first (46285bf)from vertexai.types import TypeName without needing to run from vertexai import types first (f4a6cbe)bigtable_metadata field name in FeatureOnlineStore (e0bc3d8)bigtable_metadata field name in FeatureView (e0bc3d8)enableDirectBigtableAccess field name in FeatureOnlineStore` (e0bc3d8)filter in message .google.cloud.aiplatform.v1beta1.ListSessionsRequest is changed (e0bc3d8)package_spec in message .google.cloud.aiplatform.v1.ReasoningEngineSpec is changed (e0bc3d8)package_spec in message .google.cloud.aiplatform.v1beta1.ReasoningEngineSpec is changed (e0bc3d8)ReasoningEngineSpec is changed (e0bc3d8)ReasoningEngineSpec is changed (e0bc3d8)GenAI Client(evals) - Add pass_rate to AggregatedMetricResult and calculate it for adaptive rubric metrics.
pass_rate to AggregatedMetricResult and calculate it for adaptive rubric metrics. (1f1f67e)build options in Agent Engine GCS Deployment. (28499a9)build options in Agent Engine source-based Deployment. (f7e718f)Reenable VertexAiSession for streaming_agent_run_with_events
Add telemetry enablement env for agent engines deployed using module
Default to "unspecified" for telemetry enablement
Switch tracing APIs in preview AdkApp.
cloudtrace.googleapis.com to telemetry.googleapis.com for tracing API.agent_info.agent in create_evaluation_run in Vertex AI GenAI SDK evals. (d02a7da)GenAI SDK client - Enabling new data driven prompt optimization for prompts from Android API by passing
dataset in create_evaluation_run in Vertex AI GenAI SDK evals (741c6ad)Add initial support for Python v3.14
app input in AdkApp template (10ca56f)show in Vertex AI GenAI SDK evals (d62afc3)Add STZ support for Endpoint.deploy and Model.deploy (preview).
create_evaluation_set and create_evaluation_item methods to Vertex AI GenAI SDK evals (83553a9)evaluation_df input support to create_evaluation_run method in Vertex AI GenAI SDK evals (7a1262b)inference_configs to create_evaluation_run method in Vertex AI GenAI SDK evals (5b5e6bd)labels to EvaluationRun in Vertex AI GenAI SDK evals (67cf80b)metrics to create_evaluation_run method in Vertex AI GenAI SDK evals (ce35218)show method for EvaluationRun for Vertex AI GenAI SDK evals (dc16065)remove deprecated transfer_to_agent parameter
ScaleToZeroSpec is added (62ea30a)class_methods in Agent Engine config. (af8c898)Add ability to use existing sessions for streaming_agent_run_with_events calls.
get_evaluation_set and get_evaluation_item methods to Vertex AI GenAI SDK evals (7757886)show method for EvaluationRun class in Vertex AI GenAI SDK evals (0c932b9)GenAI Client(evals) - Add EvaluationResults to get_evaluation_run method response in Vertex AI GenAI SDK evals
Mitigate breaking change in GenAI SDK for ADK on AgentEngine
create_evaluation_run method to Vertex AI GenAI SDK evals (7db9b4f)get_evaluation_run method to Vertex AI GenAI SDK evals (bbbadbc)dedicated_endpoint_disabled in message .google.cloud.aiplatform.v1beta1.DeployRequest is changed (0722687)Expose user reservation for CustomModel
Add experimental async list_prompts and list_version methods to prompt management
context in Grounding metric (e75d91f)Your coding agent can read these notes before it upgrades. Set up the MCP server →