pypi / google-cloud-aiplatform
google-cloud-aiplatform
What changed in google-cloud-aiplatform 1 to 2, from its official release notes, upgrade guides and advisories. Sources checked 2026-10-18.
2 security7 new APIs and options13 fixes and other
See which of these apply to your code, ranked, in your agent:
npx @unstale/cli initgoogle-cloud-aiplatform 220 changes
New APIs and options 7
Add agent_card to AgentRuntimeConfig for runtimes.create and update.
Agent Engine create and update requests now surface the LRO name.
Evals support cross-region routing.
Adds a method to load an EvaluationDataset from an ADK eval set.
Evals support per-step judges.
LiveClientSetup now includes labels.
The agentplatform genai SDK adds serving profiles support.
Fixes and other changes 13
Display N/A for missing summary scores in the comparison report.
Treat NaN and array cells as missing in eval DataFrames.
Send list and dict context to the evaluation service.
Raise google.genai.errors.APIError when Agent Engine create or update LROs fail.
Back off only on retryable errors during inference and agent-run retries.
agent_engines now allows opentelemetry-instrumentation-google-genai 1.x.
Pass client credentials and http_options to per-thread runtime clients.
Fix run_inference for local ADK agents and remote runtime N+1.
Save Gemini agent run_inference results to dest.
Show intermediate_events as a conversation trace in show().
Session events are replayed into a scratch session, which is deleted afterwards.
async_stream_query is non-blocking by using the async execution client.
clone() keeps the AdkApp subclass, plugins, and credential_service_builder.
google-cloud-aiplatform 12 changes
Security 2
Google Cloud Vertex AI has a a vulnerability involving predictable bucket naming
Google Cloud Vertex AI SDK affected by Stored Cross-Site Scripting (XSS)