GoogleApi.ContentWarehouse.V1.Model.KnowledgeAnswersSensitivityStoragePolicy
Table of Contents ▼
Jump to a specific part of the page:
Description
Policies controlling the storage.
Attributes List
This module has the following attributes (case-insensitive ascending order):
Attributes
-
encryptArgumentValue
(type:boolean()
, default:nil
)
- The contents of the argument value should be encrypted before being written to a persistent storage (even if the storage has short time-to-live). No-op when specified at intent level. -
encryptFieldCandidateValueTerm
(type:boolean()
, default:nil
)
- If true, triggers encryption of |value_term| [1] and move into |sensitive_value_term|. This is useful to support finer grain encryption controls for disambiguation, where the broader |encrypt_argument_value| is too restrictive. In particular, when we know that the values being disambiguated are non-sensitive (e.g. opaque event IDs), but the value_terms are sensitive (e.g. event titles), it can be helpful for debugging purposes to circulate the value unencrypted. This policy only has an effect when configured at argument-level. It has no effect in Intent or PreviousQuery stanzas. See go/delete-event-disambiguation-redaction-v2 for context and usage. [1]cs/s:quality.dialog_manager.IntentStateSignals.FieldCandidate.value_term -
encryptQueryAnnotationData
(type:boolean()
, default:nil
)
- If this is set true, we encrypt QueryAnnotationDataProto prior to writing it to Footprint ASSISTANT_EPHEMERAL corpus. This enables restricting ACL to the data. -
scrubAuxiliaryFieldsInConversationSnapshot
(type:boolean()
, default:nil
)
- If this is set to true, the following fields in ConversationSnapshot are scrubbed: circulated_state.squery spoken_query * All client_op arguments |from_assistant| interactions
Type
Function
@spec decode(struct(), keyword()) :: struct()Data sourced from HexDocs : GoogleApi.ContentWarehouse.V1.Model.KnowledgeAnswersSensitivityStoragePolicy