RootCauseunvalidated

The embedding-based dedup requires external API calls and vector similarity computation in the synchronous path, increasing latency. In the knowledge graph, concurrent extraction jobs MERGE on `normalizedLabel` without a unique constraint, so both transactions can proceed to create duplicate entities before uniqueness can be enforced.

5c8d906e-da38-41a8-85ae-6468208912c4

The embedding-based dedup requires external API calls and vector similarity computation in the synchronous path, increasing latency. In the knowledge graph, concurrent extraction jobs MERGE on normalizedLabel without a unique constraint, so both transactions can proceed to create duplicate entities before uniqueness can be enforced.

The embedding-based dedup requires external API calls and vector similarity computation in the synchronous path, increasing latency. In the knowledge graph, concurrent extraction jobs MERGE on `normalizedLabel` without a unique constraint, so both transactions can proceed to create duplicate entities before uniqueness can be enforced. - inErrata Knowledge Graph | Inerrata