Discovery API monitoring and real-time diff analysis for Vertex AI Agent Builder.
Never miss an API update, method deprecation, or breaking change in Vertex AI Agent Builder.
Discovery Engine introduces new APIs for managing UserAvailability resources, including batch updates and listing. However, this update includes breaking changes: `name` and `parent` parameters are now strictly required in certain contexts, and `verification` and `updateTime` properties are now read-only.
A breaking change makes the `verification` field in `EngineFeaturePolicyMember` read-only. Additionally, Discovery Engine now supports granular per-feature access policies on `Engine` resources, allowing you to control which users or groups can utilize specific engine features.
This update introduces granular feature access policies for Discovery Engine instances, allowing you to control who can use specific features. A breaking change makes the `verification` field in `EngineFeaturePolicyMember` read-only.
This update introduces breaking changes by making several fields read-only across Engine, Project, and AssistAnswer schemas. Developers must update any code attempting to write to these properties. Additionally, new capabilities include customer-managed OAuth scopes for connectors, global feature toggle controls for engines, and the ability to specify provisioning intent for projects.
Discovery Engine introduces new controls for global feature rollout stages and provides read-only access to effective feature states. However, this update includes significant breaking changes as many properties related to engine configuration, project experience, and tool display names have been made read-only.
This update introduces several breaking changes where numerous API properties have been made read-only. Developers must update their code if they were previously writing to these fields. Additionally, new capabilities for managing feature toggles, provisioning project experiences, and custom OAuth scopes are now available.
Discovery Engine now offers more granular feedback options, including specific reasons for positive responses and a dedicated structure for disputing Model Armor verdicts. New schemas also hint at future fine-grained control over tool visibility.
Discovery Engine now supports detailed feedback on Model Armor verdicts and response quality. New schemas also introduce fine-grained visibility controls for internal tools and allow specifying Identity Provider configurations during project provisioning.
Discovery Engine now allows developers to capture detailed user disputes for Model Armor verdicts, enabling better refinement of AI safety features. Additionally, new positive feedback reasons provide more nuanced insights into user satisfaction.
Discovery Engine now offers granular control over federated searches by allowing you to restrict data stores to specific notebooks. Additionally, new options for partial and prefix field matching are available, and conversational feedback can now include more detailed context like step sequence numbers and trajectory IDs.
This update includes significant breaking changes: the recently added `toolCalls.operations.get` and `collaborativeProjects.knowledgeSpaces.operations.get` methods have been removed. Developers should update their integrations. New features include private network OAuth configuration, notebook-level data store scoping, and enhanced field matching options.
Discovery Engine gains new capabilities for federated search data stores, allowing restriction to specific notebooks. Developers can now configure partial and prefix matching for fields, and feedback tracking gets more granular with new conversation info.
Discovery Engine now allows linking conversation feedback to specific Sobi agent tasks via a new `taskName` field. Minor internal scope adjustments were also made for `completeQuery` methods.
Discovery Engine v1alpha introduces breaking changes: the `autoModelId` field in widget configurations is now read-only, and the `name` parameter for the new `toolCalls.operations.get` method is strictly required. Additionally, a new `get` method for tool call operations and a `taskName` field for conversation feedback are now available.
A breaking change makes the `autoModelId` property in `WidgetConfigUiSettingsModelConfigInfo` read-only. Additionally, `FeedbackConversationInfo` now includes `taskName` for tracking Sobi long-running agent tasks, and `completeQuery` methods see scope reordering.
The `connectorDisplayNames` property in both `GoogleCloudDiscoveryengineV1alphaAssistAnswer` and `GoogleCloudDiscoveryengineV1betaAssistAnswer` schemas has been made read-only. This is a breaking change for applications that may have been writing to this field.
A breaking change makes `connectorDisplayNames` in `AssistAnswer` read-only, requiring updates if you were writing to it. Additionally, a new `basicAuthenticationEnabled` field provides more flexibility for OAuth integrations.
The `connectorDisplayNames` property in the `AssistAnswer` schema (v1 and v1alpha) is now `readOnly`. This is a breaking change for developers who were previously writing to this field, which provides human-readable names for internal connector agents.
Adds BillingAccountLicenseConfig schema properties and refines parameter documentation for suggestions, searches, and user events.
Introduces operations get method for collaborative project knowledge spaces, updates agent sharing configuration scopes, and adds widget and billing license properties.
Updates Discovery Engine v1 schemas, introducing read-only tag properties to WidgetConfigCollectionComponent and adding BillingAccountLicenseConfig.
Discovery Engine adds resource-level IAM filtering for Gemini Enterprise and dynamic search-as-you-type scoring overrides. Alpha users also gain support for file attachments in long-running agent tasks.
This update introduces resource-level access controls for Gemini Enterprise, per-request Search-As-You-Type overrides, and a breaking change to widget model configurations.
Discovery Engine adds granular scoring controls for Search-as-you-type and resource-level IAM filtering for Gemini Enterprise. A breaking change makes the widget 'family' field read-only.
The NotebookLM alpha API surface has been reinstated after a recent removal, now featuring stricter parameter requirements and immutable fields.
Vertex AI Agent Builder adds native Sensitive Data Protection integration for data stores and connectors, alongside new session management controls and model 'thinking' levels.
This update includes a massive breaking change removing the NotebookLM alpha API surface while introducing new Sensitive Data Protection integrations and session management controls.
Discovery Engine introduces Sensitive Data Protection (DLP) integration, configurable session management for Engines, and a breaking change making icon URLs read-only.
