DICE API Documentation/com.embabel.dice.storage/ExtractionRunSchema/MAX_CONTEXT_ID_LENGTH

MAX_CONTEXT_ID_LENGTH

const val MAX_CONTEXT_ID_LENGTH: Int = 1024

The longest tenant id this store will write.

ContextId accepts any non-blank string, and the tenant is the leading property of every key and index here. A tenant id of unbounded length is an index entry of unbounded length, and Neo4j fails that write with an index-key-size error — mid-extraction, with a message about bytes rather than about the tenant. The cap turns that into an argument rejection at the store's boundary, named, before anything is written.

1024 characters matches com.embabel.dice.proposition.extraction.ExtractionRunLimits.MAX_SOURCE_KEY_LENGTH, the most generous bound the run model already puts on a key-like string. It is far above any real tenant id and comfortably below the index key limit even once the rest of a composite key is added.

The cap is applied on writes only. A read for a tenant longer than this matches nothing, because nothing that long was ever stored — which is the fail-closed answer, and the only one a read could give.