[renovate-bot]

Note

This PR body was truncated due to platform limits.

This PR contains the following updates:

Package	Change	Age	Confidence
google-cloud-aiplatform	==1.157.0 → ==2.9.0
google-cloud-aiplatform	==2.6.0 → ==2.9.0
google-cloud-aiplatform	==1.163.0 → ==2.9.0
google-cloud-aiplatform	==1.161.0 → ==2.9.0
google-cloud-aiplatform	==1.158.0 → ==2.9.0
google-cloud-aiplatform	==2.2.0 → ==2.9.0

---

Release Notes

googleapis/python-aiplatform (google-cloud-aiplatform)

v2.9.0

Compare Source

Features

• Add missing enum values to FinishReason and ReinforcementTuningThinkingLevel (73a1a09)
• Update discovery doc (9b3f296)

Bug Fixes

• Allow litellm<1.110.0 in evaluation extra, excluding 1.107.0 (10814f7)
• Migrate usage of mock to unittest.mock (dcf3026)
• Remove brittle client-side installation_scripts path validation in Agent Engine SDK (74d17cc)
• Support per-call timeout in Agent Engine A2A operations (d309536)

Documentation

• Update AdkApp.async_stream_query docstring for multimodal message dicts and add unit tests. (1b1f8fc)

v2.8.0

Compare Source

Features

• Add service_account parameter to Model Garden deploy methods. (8779d73)
• GenAI SDK client(evals) - Add StepBasedLLMMetric for multi-step LLM-judge metrics over agent traces. (954c90d)
• GenAI SDK client(evals) - Support StepBasedLLMMetric in create_evaluation_metric. (84df3a7)
• Support git_source in client.runtimes for deploying from public Git repositories (affc088)
• Update discovery doc (720b112)

Bug Fixes

• Cap sentencepiece below 0.2.3 in the tokenization extras (bb5ff02)
• Fail open on any error resolving the project ID in the ADK framework. (2ec72e8)
• GenAI Client(evals) - Handle agent traces that end without a text event in run_inference (d62419a)
• Warn when the pipeline output bucket is not owned by the project, and add aiplatform.init(enforce_bucket_ownership=True) to raise an error instead (27ee0dc)

v2.7.0

Compare Source

Features

• Add support for Python 3.15 (9b9f2b3)

Bug Fixes

• Build the VPC-SC psc_endpoint URL as http://{psc_endpoint}:{port} in Sandboxes.send_command and ws://{psc_endpoint}:{port} in Sandboxes.generate_browser_ws_headers, instead of https://{psc_endpoint} on 443. The sandbox router pod serves plain HTTP on the caller-supplied port (default 8080) and does not terminate TLS on 443, so the prior URLs hung on ConnectTimeout. Non-VPC-SC branches are unaffected. (dd39926)
• GenAI Evaluation: Lazily import scikit-learn in preview autorater_utils so importing vertexai.preview.evaluation doesn't require it (d6f690e)
• Model Garden deploy_publisher_model now honors dedicated_endpoint_disabled and spot without a machine type (5b3dba9)
• Model Garden list_deployable_models, list_models and list_publisher_model_deploy_options now read the catalog from us-central1 for clients in any location (1ba7a41)
• Preserve caller session_id as agent_engine:upstream_session_id on fallback session creation in AdkApp. (f1944ac)
• Resolve agentplatform metric objects in vertexai evals (46f4e55)
• Stop deserializing the staged agent from Cloud Storage during Agent Engine deployment (d02310f)

v2.6.0

Compare Source

⚠ BREAKING CHANGES

• An existing value DEPRECATED is removed from enum State

Features

• Add ARCHIVED state to ReasoningEngineRuntimeRevision (faed3f4)
• Add client.memory_banks.update method. (0ac315e)
• Add context to Memory, MemoryRevision, and IntermediateExtractedMemory in Vertex AI v1beta1 (faed3f4)
• Add extra_context to EvaluationItemRequest (c0a8de2)
• Export ADK agent engine template logs to the Telemetry API (448abba)
• Onboard Vertex Model Garden to GenAI Python SDK: Add deploy_custom_model (46ae3d0)
• Support batch predict in Model Garden (6b2870f)
• Update discovery doc (b60bddc)

Bug Fixes

• An existing value DEPRECATED is removed from enum State (faed3f4)
• GenAI Client(evals) - Default missing rubric verdicts to False (93c52a6)
• GenAI Client(evals) - Honor per-row tool_config in Gemini inference (48ce2ae)
• GenAI Client(evals) - Keep eval_case_index aligned with dataset rows (19b0c40)
• GenAI Client(evals) - Log serialization errors in show() instead of raising AttributeError (4b1ace9)
• GenAI Client(evals) - Parse JSON-string and OpenAI-style conversation_history (1d93955)
• GenAI Client(evals) - Serialize DataFrame-backed evaluation datasets to JSON (9e5c721)

Documentation

• A comment for field create_time in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)
• A comment for field description in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)
• A comment for field display_name in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)
• A comment for field expire_time in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)
• A comment for field fact in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)
• A comment for field name in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)
• A comment for field scope in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)
• A comment for field ttl in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)
• A comment for field update_time in message .google.cloud.aiplatform.v1beta1.Memory is changed (faed3f4)

Miscellaneous Chores

• Release 2.6.0 (ed98faa)

v2.5.0

Compare Source

Features

• agentplatform: Add agent_card to AgentRuntimeConfig for runtimes.create and update (527b6dc)

Bug Fixes

• agentplatform: Allow opentelemetry-instrumentation-google-genai 1.x in agent_engines (7f08bd6)
• GenAI Client(evals) - Back off only on retryable errors in inference and agent-run retries (f6bf219)
• GenAI Client(evals) - Fix run_inference for local ADK agents and remote runtime N+1 (738e2b1)
• GenAI Client(evals) - Mirror the inference retry fix to agentplatform/private and vertexai (efa3980)
• GenAI Client(evals) - Pass client credentials and http_options to per-thread runtime clients (58c34ea)
• GenAI Client(evals) - Save Gemini agent run_inference results to dest (be0eb9a)
• GenAI Client(evals) - Send list and dict context to the evaluation service (2c2d69e)
• GenAI Client(evals) - Show intermediate_events as a conversation trace in show() (4d4ac8a)
• GenAI Client(evals) - Show N/A for missing summary scores in the comparison report (be4ea51)
• GenAI Client(evals) - Treat NaN and array cells as missing in eval DataFrames (b28737e)
• Raise google.genai.errors.APIError when Agent Engine create or update LROs fail (e631a16)

Documentation

• Update gemini model in docstrings for RAG Engine (16c3fb6)
• Update generated SDK docstring examples to use gemini-flash-latest (8010cbd)

v2.4.0

Compare Source

Features

• Add per-step judges and cross-region routing to evals (5fa1455)
• GenAI Client(evals) - add EvaluationDataset.load_from_adk_eval_set (4c045e1)
• Surface the LRO name in Agent Engine create and update requests (b1378bb)
• Update discovery doc (39693a9)

Bug Fixes

• Replay session_events into a scratch session and delete it afterwards (3647753)

v2.3.0

Compare Source

Features

• Add serving profiles support to agentplatform genai sdk (6da2b9b)
• Include labels for LiveClientSetup (6b29d01)

Bug Fixes

• Keep AdkApp subclass, plugins and credential_service_builder in clone() (cfb44c4)
• Make async_stream_query non-blocking by using the async execution client (d62114d)

v2.2.0

Compare Source

Features

• Update discovery doc (6ad6fe8)

Bug Fixes

• Allow unknown union variants when validating gaos responses in evals (a9c4e6e)
• Resolve vertexai.types against agentplatform.types (5af2f8e)
• Resolve vertexai.types against agentplatform.types (36a47bd)
• Resolve vertexai.types against agentplatform.types (75ecb3b)

v2.1.3

Compare Source

Features

• Add 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)
• Add build configuration support to Vertex AI Reasoning Engine (bea1abd)
• Add enable_zero_data_retention to Tool.ParallelAiSearch (bea1abd)
• Add enable_zero_data_retention to Tool.ParallelAiSearch (bea1abd)

Bug Fixes

• Fail open on any error resolving the project ID in the ADK templates. (30bf665)
• Fix Sandboxes.send_command failing to reach the sandbox data plane. (2b28caf)
• Import pandas lazily in the Gen AI SDK _gcs_utils module. (7c94277), refs #​5928 #​7129

Miscellaneous Chores

• Release 2.1.3 (2c7a22d)

v2.1.0

Compare Source

Features

• Add execute_bash to the Agent Engine sandbox SDK. (505bd42)
• Allow setting ingress_control_config when creating a SandboxEnvironmentTemplate via the SDK (Python, Java, JS). (2dccde8)
• Expose SandboxEnvironmentTemplateState as a standalone enum class in the Python, Java, and JS SDKs. (a31d2ba)
• Expose STATE_PAUSED, STATE_PAUSING, and STATE_RESUMING values in the SandboxState enum for SandboxEnvironment across Python, Java, and JS SDKs. (ec342d9)

Bug Fixes

• Guard the post-LRO refresh in sandboxes.pause() and sandboxes.resume() on the presence of a resource name in the LRO response. (2c60077)
• Preserve rrf_ranking_alpha=0.0 in MatchingEngineIndexEndpoint hybrid queries (7e1c95a)

Documentation

• Update CHANGELOG.md to mark version 2.0.0 as yanked (75f9361)

v2.0.1

Compare Source

⚠ BREAKING CHANGES

• agentplatform: agent_engines module renamed to runtimes.
• agentplatform: a2a tasks module is removed.
• agentplatform: sessions and sandboxes moved to top-level modules.
• agentplatform: Removed global initializer dependency. Agent frameworks now read project and location from environment variables.
• agentplatform: runtime templates import changed from agentplatform.agent_engines.templates to agentplatform.frameworks.
• agentplatform: evals.run_inference(agent=...) now accepts types.Runtime instead of types.AgentEngine.
• agentplatform: Migrate Memory Bank methods to client.memory_banks sub-module.

Features

• Add sandboxes.pause() and sandboxes.resume() methods to the Agent Engine sandbox SDK across Python, Java, and JS SDKs. (9b92c88)
• Add Shell Sandbox environment support to the Agent Engine sandbox SDK (8f87082)
• agentplatform: Support client.memory_banks.get (6dfe678)
• agentplatform: Support client.memory_banks.list (6dfe678)
• agentplatform: Support ExampleStores module for few-shot example management (5248240)
• agentplatform: Support Memory Bank configs. (6dfe678)
• Decouple agentplatform from google-cloud-aiplatform and package it as a standalone PyPI package google-cloud-agentplatform. (e0e32c2)
• Expose ingress_control_config field on SandboxEnvironmentTemplate across Python, Java, and JS SDKs. (c08ed98)
• Expose service_attachment field on SandboxEnvironmentConnectionInfo across Python, Java, JS, and Go SDKs. (b781b0e)

Bug Fixes

• Add identity_type and service_account support to Reasoning Engines and Agent Engines. (a8536aa)
• Propagate the underlying API error from AdkApp's synchronous session methods and streaming agent runs (91cc8ce)
• Serve public A2A agent card at /a2a/card on Agent Runtime (85cba30)
• Treat forwarded OAuth access tokens as ephemeral session state (7e95ce6)

v2.0.0

Compare Source

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.

v1.165.1

Compare Source

Bug Fixes

• Truncate default pipeline and staging bucket names to 63 chars (03edd90)

v1.165.0

Compare Source

Features

• Add enable_data_retention to ToolParallelAiSearch, Add step_count to ReinforcementTuningHyperParameters, Add BidiGenerateContentSetup (09acba7)
• Expose public update() and update_async() methods on Sessions and AsyncSessions client modules. (5df3113)
• GenAI Client(evals) - add import_evaluation_set (eed0b4f)
• GenAI Client(evals) - add list_evaluation_sets and delete_evaluation_set (6ee088b)
• Onboard Vertex Model Garden to GenAI Python SDK: Add deploy_publisher_model support (9ef9093)

Bug Fixes

• Check GCS ownership for default pipeline bucket (2baea5b)
• Make the Telemetry API enablement check non-fatal in AdkApp.set_up(). (20205a8)
• Parse Server-Sent Events in Agent Engine streaming responses (7cb2fe9)
• Stop logging the "experimental" warning when creating a computer-use sandbox in the Vertex AI GenAI Python SDK. (6056aa0)

v1.164.0

Compare Source

Features

• Add build configuration support to Vertex AI Reasoning Engine (e20c003)
• Add encryption_spec (CMEK) support to EvaluationMetric SDK. (6dd3179)
• Add encryption_spec (CMEK) support to EvaluationRun SDK (50a43af)
• Add encryption_spec (CMEK) support to EvaluationSet resource (c2244f9)
• Add endpoint field to BatchPredictionJob for bring-your-own-endpoint batch prediction (19144a4)
• Add sandbox templates and snapshots submodules to the Vertex AI GenAI SDK for Java. (f7e3a70)
• Add Tool.ExaAiSearch to Vertex AI v1 (19144a4)
• Add Tool.ExaAiSearch to Vertex AI v1beta1 (19144a4)
• Allow users to configure max wait time for prompt management queries for Python and JS (ad63bfe)
• Allow users to search for spaces without requiring admin privileges (19144a4)
• Forward per-request labels to RunConfig in streaming_agent_run_with_events (f8eb68c)
• GenAI Client(evals) - async auto-create EvaluationExperiment parity (e689f0e)
• GenAI Client(evals) - auto-create EvaluationExperiment in create_evaluation_run (eb75e14)
• GenAI Client(evals) - Pin SDK predefined metrics to latest Gemini 3.5 Flash versions by default (f5d94b4)
• GenAI Client(evals) - Support show() for EvaluationDataset built from eval_cases (5b1a0aa)
• Support Endpoints module for OpenModel prediction, and .get/undeploy/delete (8848353)
• Update Compute v1 to revision 2026072 (19144a4)

Bug Fixes

• Add aiohttp to ADK extra requirements. (390f2d1)
• deps: Update the Java code generator (gapic-generator-java) to v2.75.0 (19144a4)
• Enable automatic session creation in ADK template. (f343270)
• GenAI Client(evals) - skip agent config fetch when agent and client locations differ (8b43add)
• Import iam_credentials_v1 lazily in the sandboxes module. (2a18bff)
• Persist the AgentCard on Agent Engine update() (5d5a1ca), refs #​7064
• Remove the google-cloud-iam dependency from the agent_engines extra. (25deb44)
• Update generator configuration and SDK bindings for Agent Engine Session update calls to return Session directly. (2702857)

Documentation

• A comment for field order_by in message .google.chat.v1.SearchSpacesRequest is changed (19144a4)
• A comment for field query in message .google.chat.v1.SearchSpacesRequest is changed (19144a4)
• A comment for field spaces in message .google.chat.v1.SearchSpacesResponse is changed (19144a4)
• A comment for field use_admin_access in message .google.chat.v1.SearchSpacesRequest is changed (19144a4)
• A comment for method SearchSpaces in service ChatService is changed (19144a4)
• Communicate on upcoming updates in the next major version (c64b25f)
• Communicate on upcoming updates in the next major version (76b9616)
• Update copyright year (19144a4)
• Update various comments (19144a4)

v1.163.0

Compare Source

Features

• GenAI Client(evals) - add create EvaluationExperiment (07ade2e)
• GenAI Client(evals) - add delete/update EvaluationExperiment (09e5a65)
• GenAI Client(evals) - add get/list EvaluationExperiment (52b1c24)
• Onboard Vertex Model Garden to GenAI Python SDK: Add export_open_model support (c2ba5d2)
• Support us mREP endpoint for Batch Prediction, Training, and Online Prediction. (96a9fcf)

Bug Fixes

• Depend on the autogen distribution instead of ag2 for the AG2 integration. (dbcf815)
• GenAI client - Resolve opentelemetry-resourcedetector-gcp dependency error (a16fc0d)
• Merge sandbox-only turns in SDK-side interaction-to-AgentData conversion. (f93c455)
• Send display_name when creating a sandbox environment template. (243ca19)

v1.162.0

Compare Source

Features

• Add detailed agent tool definitions for display (762079f)
• Add Route to GroundingChunk.Maps in Vertex AI v1 (4331d55)
• Add Route to GroundingChunk.Maps in Vertex AI v1beta1 (4331d55)
• Delegate agent resolution to server via geminiAgentConfig (ab757c7)
• GenAI client - Add warning on vertexai.Client instantiation and recommend agentplatform.Client (f141415)
• GenAI Client(evals) - multi-turn user simulation for Gemini Agents API run_inference (c74e996)

Bug Fixes

• Use the correct enum for RagFile.file_status (23145f3)

Documentation

• Expand GroundingChunk.Maps documentation to describe route support (4331d55)
• Expand GroundingChunk.Maps documentation to describe route support (4331d55)
• GenAI client - Add example usage to RAG docstrings (df3dcd9)
• Update product name from "Vertex AI" to "Gemini Enterprise Agent Platform" in README, and add a note to rename Agent Engine to Agent Runtime in upcoming major release (a3179c2)

v1.161.0

Compare Source

Features

• Add FeedbackService to Agent Platform Python SDK (f6ecd67)
• Add identity_type to Reasoning Engine public protos (66d4837)
• Add identity_type to Reasoning Engine public protos (66d4837)
• Add support for FeedbackContext to Agent Platform Python SDK (f5f750c)
• Expose retrieval_queries in GroundingMetadata for the v1 API (66d4837)
• GenAI Client(evals) - evaluate Gemini Agents API agents via interaction (a44c520)
• GenAI Client(evals) - evaluate() agent parameter for interaction-id datasets (135224c)
• GenAI Client(evals) - exponential backoff for Gemini agent interaction polling (1f3c074)
• GenAI Client(evals) - generate_conversation_scenarios support for Gemini Agents API agents (df4e672)
• GenAI Client(evals) - run_inference support for Gemini Agents API agents (852c546)
• Onboard Vertex Model Garden to GenAI Python SDK: Add list_custom_model_deploy_options support (239bc13)
• Resolve Interactions API data for agent trace display in eval results. (f158517)
• Support creating evaluation runs from pre-existing Interactions API data (b430014)
• Support Prompt Management in the JS GenAI Modules (89d2b91)

Bug Fixes

• Remove OpenTelemetry event logger setup since the Events API/SDK has been deprecated and is soon being removed entirely. (73fde8d)

Documentation

• Add result_parsing_function field to LLMMetric SDK types (8f1a0a1)

v1.160.0

Compare Source

Features

• Add revision_labels, revision_expire_time, revision_ttl, disable_memory_revisions, metadata, metadata_merge_strategy and overlap_event_count to Memory Bank IngestEvents API (e169af9)
• GenAI Client(evals) - add Gemini Agents API related eval types (c8a38a0)
• Onboard Vertex Model Garden to GenAI Python SDK: Add list_publisher_model_deploy_options support (ce463e6)

Bug Fixes

• Add mTLS and telemetry endpoint configurations to preview AdkApp (c3c1836)
• Avoid shadowing of built-in list in Python 3.14 (195ca27)
• Honor explicitly passed api_endpoint when location is "global" (f9e3c6e), closes #​6958

v1.159.0

Compare Source

⚠ BREAKING CHANGES

• 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.

Features

• Add data_sharing_enabled_provider field (ModelProvider enum) to PublisherModelConfig (51c4d78)
• Add deprecation warning to vertexai.rag module in favor of agentplatform.Client().rag (9ffcbce)
• Add support for session TTL and expiration in Vertex AI session service (a2af300)
• GenAI client - Add upload_file method to RAG module (13b685b)
• GenAI client - Log operation name when RAG operation fails (e368858)
• GenAI SDK client(multimodal) - Return the assembled table as a DataFrame from datasets.assemble. (8e52bd2)
• Onboard Vertex Model Garden to GenAI Python SDK: Add list_models and list_deployable_models support (9cd63ab)

Bug Fixes

• Added safety checks in download_from_gcs (2f5c8f8)
• ai_platform: Black check (97ee94a)
• aiplatform: Safely parse bq:// URIs using removeprefix (97ee94a)
• feature_store: Safely parse bq:// URIs using removeprefix (97ee94a)
• Fix the async_stream_query method. Restore yield_parsed_json method lost during vertex -> agentplatform

[!IMPORTANT]
✂ PR body was truncated to here.

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Never, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about these updates again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.

[gemini-code-assist]

Code Review

This pull request updates the google-cloud-aiplatform dependency version to 2.9.0 across multiple requirements.txt and requirements-test.txt files in various project directories. I have no feedback to provide as there are no review comments.