Asopitech VS
Elasticsearch vs Alopex DB
Elasticsearch's kNN vector search and refresh behavior. Also explains how it differs from Alopex DB.
Updated: Oct 12, 2026
What Elasticsearch is
Elasticsearch is a search engine. Alongside full-text search, it has kNN vector search.
How it works
- Vectors: kNN search runs on dense_vector fields.
- Search types: approximate kNN for production and exact brute-force kNN.
- Embedding generation: semantic_text and inference services create embeddings.
- Refresh: writes become visible to search after a refresh. The default refresh_interval is 1 second, and 5 seconds on Serverless.
- OpenSearch k-NN: in OpenSearch, the k-NN plugin (Apache-2.0) provides approximate kNN search. The engine is Faiss, Lucene, or NMSLIB, and the methods are HNSW and IVF. JVector, with a DiskANN-style method, is a separate plugin.
Main features
Full-text search and vector search run in one search engine. Embedding generation also runs inside Elasticsearch.
Limits
- Writes become visible to search only after a refresh.
- The documentation says refresh=true for immediate visibility creates small segments that cost indexing, search, and merge time.
- Exact kNN scores every matching document, so it does not suit large data sets.
- OpenSearch’s NMSLIB engine is deprecated.
- The OpenSearch JVector plugin cannot be installed in the same cluster as the k-NN plugin.
- With OpenSearch’s Faiss engine, filtering is applied after the search.
How it differs from Alopex DB
Both have full-text search and vector search.
Alopex DB brings SQL, vector search, and full-text search together in one engine. It runs as a single embedded file.
Alopex DB runs full-text and vector search inside SQL and keeps search next to your app. You can try it from the quickstart.