Member Junction
    Preparing search index...

    Parameters for upserting a record's Feature Values, one row per Feature Value key; every key part is required.

    interface RecordFeatureValuesParams {
        aiModelID: string;
        aiPromptRunID?: string;
        constraintHash?: string;
        contextUser: UserInfo;
        entityID: string;
        featureValueCacheID?: string;
        historyOnly?: boolean;
        inputHash: string;
        outputs: FeatureValueOutputItem[];
        parametersHash: string;
        pipelineVersion: number;
        processRunDetailID?: string;
        processRunID?: string;
        promptID?: string;
        promptVersionHash: string;
        provider: IMetadataProvider;
        recordID: string;
        recordProcessID: string;
        source: "Human" | "Pipeline";
    }
    Index

    Properties

    aiModelID: string

    The pin's primary AI model; null when the prompt has no Active model, or for a human value.

    aiPromptRunID?: string
    constraintHash?: string
    contextUser: UserInfo
    entityID: string
    featureValueCacheID?: string
    historyOnly?: boolean

    The rows are history (a decision an escalation superseded), never made the feature's latest value again.

    inputHash: string

    SHA-256 of the record's normalised inputs (the cache key hash); null when there is none.

    parametersHash: string

    The pin's parameters hash; null for a human value.

    pipelineVersion: number

    The pipeline version that computed the values; null when unversioned.

    processRunDetailID?: string
    processRunID?: string
    promptID?: string
    promptVersionHash: string

    The pin's prompt version hash; null for a value no prompt computed.

    The caller's provider (on the server, the request's own connection); the rows are read and written through it.

    recordID: string
    recordProcessID: string
    source: "Human" | "Pipeline"