OptionalDefaultGlobal default truncation limit (characters) for large string fields. Default 1000.
OptionalFieldsPer-field overrides keyed by ContentItem field name (see MJContentSourceEntity_IContentSourceVectorMetadataFieldConfig).
OptionalFieldWhich ContentItem fields go into metadata. Mirrors the entity pipeline's field strategy. When UNSET, the standard curated content set is used (the historical default): the identity keys + ContentSourceID / ContentSourceTypeID + Title / Description / URL + Tags. When set:
Included: true in Fields (explicit
inclusion wins over the eligibility heuristics — a uniqueidentifier / PK / _mj* field
can be included by name; only genuinely unstorable binary types are refused).Included: false in Fields.Fields — no system keys except Entity (always kept so
content search results stay correctly labeled), and the toggles flip to opt-in (default
false). Keeps metadata minimal. NOTE: under 'explicit' a search hit's record id is
recoverable only when VectorIDStrategy='recordId' (the default), where the vector's own id
is the chunk id; with 'hash' the id would need to be kept explicitly.OptionalIncludeInclude the content entity's icon. Default true under a set strategy; opt-in under 'explicit'.
OptionalIncludeInclude the item's Tags array. Default true (and under the curated default); opt-in under 'explicit'.
OptionalIncludeWhen true, include the embedded text in metadata under the 'Text' key (which surfaces as the search snippet). Default false — external hydrators read the authoritative text from the ContentItem / ContentItemChunk row, so the copy is usually unnecessary storage. Honored under every strategy (including the curated default).
OptionalIncludeInclude __mj_UpdatedAt for recency sorting. Default true under a set strategy; opt-in under 'explicit'.
Controls which keys land in a content vector's metadata. Vector metadata is expensive in a vector database (storage + query performance). This mirrors the entity-vectorization pipeline's metadata-control structure (field strategy, per-field overrides, storage-type coercion, truncation, opt-out toggles), adapted to content-item vectors.
Field values are read from the parent ContentItem. The identity keys (Entity + RecordID, plus ContentItemID / Sequence for chunk vectors) are managed as system keys — see FieldStrategy.