Canonical representation of query sensitivites. See go/sensitive-intents for more details.

Attributes

• accountProvenance (type: list(GoogleApi.ContentWarehouse.V1.Model.QualityQrewriteAccountProvenance.t), default: nil) - Used to annotate the provenace of cross-account personal data. See go/cross-account-understanding. Sensitivity could be annotated at query, intent, and argument levels. Query and intent could have data from multiple accounts, so this field is repeated. A sensitive knowledge_context.PreviousQuery might be dropped to prevent leaking cross-account data via Genie rewrite. For arguments, the best practice is to not blend multi-account data, and this field should be treated as singular to make ownership clear. When publishing attentional entities, contextual NLU might drop an argument that contains data from a different account for data protection. Mixing multi-account data in one argument will cause data from the primary account to be dropped altogether, which is an unnecessary quality loss.
• instruction (type: GoogleApi.ContentWarehouse.V1.Model.KnowledgeAnswersSensitivityInstruction.t, default: nil) - Instruction of handling sensitive intent/argument data. Can be specified in Intent Catalog.
• source (type: list(String.t), default: nil) - This should be systematically added without requiring the feature developers to add a source. This is for debug purpose as to whether the Sensitivity's trace/path included any of landmark code path helpful for tracing back the sensitivity sources. One should add a new source when one sees fits. It's chronological order as to which source is added first. One should not manually add a source. The same source can be repeated if the Sensitivity object went through the same code path twice.
• type (type: String.t, default: nil) - Sensitivity type. See the enum definition below.