Retrieval Elasticsearch Retrievers and RRF

Status: public · Confidence: medium (0.685) · Basis: verified_sources

## TL;DR

Elasticsearch retrievers and reciprocal rank fusion help agents inspect how lexical, vector, semantic, and reranking stages were combined in a search request.

## Core Explanation

Modern Elasticsearch retrieval can be expressed as a retriever tree instead of a single query. That tree can combine standard lexical search, kNN vector search, RRF, linear score combination, rescorers, pinned documents, query rules, and text similarity rerankers.

For RAG debugging, an agent should preserve the full retriever object, rank windows, size, min_score, fields, child retrievers, semantic fields, kNN parameters, and final hit order. A bad answer may come from candidate generation, fusion, or a downstream reranker, so the retriever tree is more useful evidence than a flat list of hits.

## Source-Mapped Facts

- Elasticsearch retrievers documentation says a retriever is a specification that describes top documents returned from a search. ([source](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/retrievers))
- Elasticsearch retrievers documentation says retrievers can form a tree-like structure that clarifies the order of operations during search. ([source](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/retrievers))
- Elasticsearch retrievers documentation lists standard, knn, rrf, linear, rescorer, pinned, diversify, rule, and text similarity reranker retrievers. ([source](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/retrievers))
- Elasticsearch reciprocal rank fusion documentation describes RRF as a method for combining multiple result sets with different relevance indicators. ([source](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/reciprocal-rank-fusion))
- Elasticsearch retrievers documentation says min_score filtering with compound retrievers is applied after compound scoring such as RRF or linear combination. ([source](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/retrievers))

## Further Reading

- [Elasticsearch Retrievers](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/retrievers)
- [Elasticsearch Reciprocal Rank Fusion](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/reciprocal-rank-fusion)