Transform dense_vector fields to knn_vector
Elasticsearch 7.x introduced the dense_vector field type for storing numeric vectors used in k-nearest neighbor (k-NN) and semantic search. Amazon OpenSearch Service and Amazon OpenSearch Serverless NextGen do not use that field type — they use the knn_vector field type instead. When you migrate metadata, Migration Assistant applies a built-in transformation that rewrites dense_vector mappings into equivalent knn_vector mappings on the Amazon OpenSearch Service domain or Amazon OpenSearch Serverless NextGen collection, so your vector indexes can be created on the target without manual mapping edits.
This transformation runs automatically during the metadata migration phase, alongside the other built-in field-type transformations (string to text and keyword, and flattened to flat_object). For an overview of where metadata migration fits in the workflow, see Migrate metadata.
Note
This transformation is metadata-only — it changes how the index is defined on the target. The vector values themselves are carried over during document backfill from the source snapshot. Migration Assistant does not recompute or re-embed vectors.
