Easysearch vs Elasticsearch Vector Search: Compatibility Explained in One Guide

Conclusion

Easysearch supports vector search but its API is not compatible with Elasticsearch 8.x vector API; they use different field types and query DSL.

Core Differences

Field type: Easysearch uses knn_dense_float_vector / knn_sparse_bool_vector; Elasticsearch uses dense_vector, sparse_vector, semantic_text.

Query entry: Easysearch uses query.knn_nearest_neighbors; Elasticsearch uses top‑level knn, retriever.knn or script_score.

Vector parameter name: Easysearch uses vec.values; Elasticsearch uses query_vector.

Approx/Exact control: Easysearch sets model: lsh / exact; Elasticsearch relies on approximate kNN or custom script_score.

Candidate parameter: Easysearch uses candidates; Elasticsearch uses num_candidates.

Compatibility: Easysearch ❌ not compatible with ES 8.x vector API; Elasticsearch ✅ native.

DSL Comparison

Easysearch

Index creation example:

PUT /my-vectors
{
  "mappings": {
    "properties": {
      "embedding": {
        "type": "knn_dense_float_vector",
        "knn": {
          "dims": 4,
          "model": "lsh",
          "similarity": "cosine",
          "L": 99,
          "k": 1
        }
      }
    }
  }
}

Search example:

POST /my-vectors/_search
{
  "size": 10,
  "query": {
    "knn_nearest_neighbors": {
      "field": "embedding",
      "vec": {
        "values": [0.10, -0.02, 0.87, 0.40]
      },
      "model": "lsh",
      "similarity": "cosine",
      "candidates": 100
    }
  }
}

Elasticsearch

Index creation example:

PUT /amazon-reviews
{
  "mappings": {
    "properties": {
      "review_vector": {
        "type": "dense_vector",
        "dims": 4,
        "index": true,
        "similarity": "cosine"
      }
    }
  }
}

Search example:

POST /amazon-reviews/_search
{
  "knn": {
    "field": "review_vector",
    "query_vector": [0.1, 0.2, 0.29, 0.41],
    "k": 2,
    "num_candidates": 5
  }
}

Exact Search Differences

Easysearch switches the model to exact while keeping the same query entry:

{
  "query": {
    "knn_nearest_neighbors": {
      "field": "my_vec",
      "vec": { "values": [0.1,0.2,0.3,0.4] },
      "model": "exact",
      "similarity": "cosine"
    }
  }
}

Elasticsearch uses a script_score query with a custom script:

{
  "query": {
    "script_score": {
      "query": { "match_all": {} },
      "script": {
        "source": "cosineSimilarity(params.queryVector, 'product-vector') + 1.0",
        "params": { "queryVector": [0.1,0.2,0.3,0.4] }
      }
    }
  }
}

Common Migration Pitfalls

Copying a dense_vector field directly to Easysearch causes an error; Easysearch expects knn_dense_float_vector.

Sending Elasticsearch top‑level knn query to Easysearch fails because the query tree structure differs.

Assuming functional compatibility equals API compatibility; the DSL must be rewritten, not just parameter names.

Attempting performance tuning before a minimal demo works; first verify the full index‑write‑search flow with a 4‑dim dummy dataset.

Selection Advice

If you heavily depend on the ES 8.x vector ecosystem (dense_vector, knn, semantic_text), migrating to Easysearch is a full DSL rewrite and requires careful evaluation.

For new projects willing to follow documentation, Easysearch’s docs are straightforward and can be followed without hitting the above pitfalls.