ElasticsearchVectorStore

[gikeihi]

Why is the query field "embedding" hardcoded in the ElasticsearchVectorStore class?

			SearchResponse<Document> res = this.elasticsearchClient.search(sr -> sr.index(this.options.getIndexName())
				.knn(knn -> knn.queryVector(EmbeddingUtils.toList(vectors))
					.similarity(finalThreshold)
					.k((long) searchRequest.getTopK())
					.field("embedding")
					.numCandidates((long) (1.5 * searchRequest.getTopK()))
					.filter(fl -> fl
						.queryString(qs -> qs.query(getElasticsearchQueryString(searchRequest.getFilterExpression())))))
				.size(searchRequest.getTopK()), Document.class);

[ilayaperumalg]

Hi @gikeihi , This is because the ElasticSearchVectorStore's document object has the embeddings field mapped to "embedding" and by default the index mapping is also configured to use the same field name "embedding" as in here.

[gikeihi]

@ilayaperumalg
Thank you for your reply. I’d like to know why. Why can’t this field be configured? I already have an index whose vector field is “vector.” What should I do?

[ilayaperumalg]

@gikeihi I agree this should be configurable. I will keep this issue open to track this.

[sobychacko]

Closing via #2175.