OptionaldimensionsReduced embedding dimensions from MJ: Vector Indexes.Dimensions, when set. Passed to the
embedding call so models that support it (e.g. text-embedding-3-*) produce shorter vectors.
Undefined means the model's native dimensionality (matches PR #3175's entity-path behavior).
OptionalembeddingEmbedding goes through AIEmbeddingRunner, pinned to embeddingModelID. For older callers this is still a working driver, built on first read with the legacy environment-variable key.
The embedding model (MJ: AI Models.ID). Every embedding call for this index pins it as
ModelID, so the index only ever holds vectors from this model.
OptionalproviderParsed MJ: Vector Indexes.ProviderConfig JSON — the opaque, provider-specific blob passed
to the vector DB so drivers can read their own settings (e.g. Pinecone's namespaceField).
Undefined when the index has no ProviderConfig or it is not valid JSON.
Resolved vector infrastructure for a specific (embeddingModel + vectorIndex) pair. Items sharing the same pair are batched together for efficient processing.