Member Junction
    Preparing search index...

    Hierarchy (View Summary)

    Index

    Constructors

    • Parameters

      • Optionalprovider: IMetadataProvider

        Optional metadata provider to bind this instance (and all of its metadata/view operations) to. Server-side callers servicing per-request connections MUST pass their request-scoped provider (e.g. this.ProviderToUse from a BaseEntity subclass) — relying on the global default silently uses the wrong connection in multi-provider scenarios. Falls back to the global default provider when omitted.

      Returns VectorBase

    Properties

    _currentUser: UserInfo
    _metadata: Metadata
    _provider: IMetadataProvider = null
    _runView: RunView

    Accessors

    • get Metadata(): IMetadataProvider

      Provider-aware metadata access: returns the explicit provider when one was supplied (constructor or Provider setter), otherwise the global default. All metadata operations in this class hierarchy go through this getter so a bound instance never leaks onto the global provider.

      Returns IMetadataProvider

    • get Provider(): IMetadataProvider

      Optional metadata provider override. Pass the provider to the constructor (preferred) or set instance.Provider = providerToUse before invoking helper methods in multi-provider contexts. Setting it rebinds the internal RunView instance so view execution rides the same provider. Falls back to the global default when unset.

      The fallback is the global provider itself, never the Metadata wrapper: consumers type-check what they get here (e.g. VectorDBBase.TryWireColocatedHost needs an IColocatedVectorHost), and the wrapper fails those checks.

      Returns IMetadataProvider

    • set Provider(value: IMetadataProvider): void

      Parameters

      Returns void

    Methods

    • Returns true when PageRecordsByEntityID can serve the given entity via keyset (seek) pagination — i.e. the entity has a single-column PK on a comparable type. Callers iterating large tables should consult this and pass AfterKey when true.

      Parameters

      • entityID: string | number

      Returns boolean

    • Resolve the embedding model to generate vectors with.

      Pass an id wherever one exists. The width of an embedding is a property of the model, and the width an index accepts is fixed when the index is created, so which model this returns decides whether the vectors fit. The no-id branch cannot know that and is a last resort only.

      The no-id branch used to be a bare find() over AIEngine.Instance.Models, which BaseAIEngine.Config loads with no OrderBy and no IsActive filter. So it returned whichever Embeddings row the view happened to hand back first — a different model on a different database, or after an unrelated insert, and quite possibly a deactivated one. An arbitrary answer is worse than a wrong one: the same code produces different widths in different environments and nothing in the failure names the model that was chosen.

      It is now deterministic — active models first, then highest PowerRank, then name — and it says out loud which model it picked and that nobody asked for it.

      Parameters

      • Optionalid: string

      Returns MJAIModelEntityExtended

    • Type Parameters

      Parameters

      • entityName: string
      • extraFilter: string

      Returns Promise<T>

    • Saving an Entity in any vector related package needs the CurrentUser property to be set on the entity So this is a simple wrapper to set it before saving

      Parameters

      Returns Promise<boolean>