Use knn inside functional_score query

Elastic Stack Elasticsearch
,
1

I am using elasticsearch-java client (co.elastic.clients) v8.11

I need to create a knn query inside functional_score query similar to query shared below.

{
 "size": 3,
 "query": {
   "function_score": {
     "query": {
       "knn": {
           "field": "embedding",
           "query_vector": [2,2,2,0],
           "num_candidates": 10,
           "_name": "knn_query"
         }
     },
     "functions": [
       {
         "filter": { "match": { "department": "men" } },
         "weight": 100
       },
       {
         "filter": { "match": { "department": "women" } },
         "weight": 50
       }
     ]
   }
 }
}

I have created a KnnQuery object,

KnnQuery knnQuery = new KnnQuery.Builder()
                .field("title_vector")
                .queryVector(embeddings)
                .numCandidates(10)
                .similarity(similarityScore)
                .boost(getKnnBoost())
                .filter(filterQuery._toQuery())
                .build();

Now, how can I use it inside my FunctionScoreQuery ?

2

Hello and welcome! You can add any query to FunctionScoreQuery by using its builder like so:

FunctionScoreQuery.of(fs -> fs
    .query(q -> q.knn(your-knn-query))
    .functions(list-of-functions));
3

Welcome!

I'm not sure it's doable with this version as it has been introduced as a query in 8.12. Please update to 8.15.0 and you will be probably able to do something like:

client.search(sr -> sr
        .index(indexName)
        .query(q -> q.functionScore(
                fsq -> fsq
                        .query(qknn -> qknn.knn(
                                k -> k.field("vector").queryVector(0.9f, 0.4f, 0.8f)
                        ))
                        .functions(fs -> fs.randomScore(rs -> rs))
        ))
, Void.class);

I added an example here:

2 Likes