Package dev.langchain4j.rag.content.aggregator

Class DefaultContentAggregator

java.lang.Object

dev.langchain4j.rag.content.aggregator.DefaultContentAggregator

All Implemented Interfaces:

ContentAggregator

public class DefaultContentAggregator
extends Object
implements ContentAggregator

Default implementation of ContentAggregator intended to be suitable for the majority of use cases.

It's important to note that while efforts will be made to avoid breaking changes, the default behavior of this class may be updated in the future if it's found that the current behavior does not adequately serve the majority of use cases. Such changes would be made to benefit both current and future users.

This implementation employs Reciprocal Rank Fusion (see ReciprocalRankFuser) in two stages to aggregate all Collection<List<Content>> into a single List<Content>. The Contents in both the input and output lists are expected to be sorted by relevance, with the most relevant Contents at the beginning of the List<Content>.
Stage 1: For each Query, all List<Content> retrieved with that Query are merged into a single List<Content>.
Stage 2: All List<Content> (results from stage 1) are merged into a single List<Content>.

Example:
Input (query -> multiple lists with ranked contents):

 home animals -> [cat, dog, hamster], [cat, parrot]
 domestic animals -> [dog, horse], [cat]
 

After stage 1 (query -> single list with ranked contents):

 home animals -> [cat, dog, parrot, hamster]
 domestic animals -> [dog, cat, horse]
 

After stage 2 (single list with ranked contents):

 [cat, dog, parrot, horse, hamster]
 

See Also:

• ReciprocalRankFuser
• ReRankingContentAggregator

Constructor Summary

Constructors

Constructor	Description
DefaultContentAggregator()

Method Summary

Modifier and Type	Method	Description
List<Content>	aggregate(Map<Query,Collection<List<Content>>> queryToContents)	Aggregates all Contents retrieved by all ContentRetrievers using all Querys.
protected Map<Query,List<Content>>	fuse(Map<Query,Collection<List<Content>>> queryToContents)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

DefaultContentAggregator

public DefaultContentAggregator()

Method Details

aggregate

public List<Content> aggregate(Map<Query,Collection<List<Content>>> queryToContents)

Description copied from interface: ContentAggregator

Aggregates all Contents retrieved by all ContentRetrievers using all Querys. The Contents, both on input and output, are sorted by relevance, with the most relevant Contents appearing at the beginning of List<Content>.

Specified by:

aggregate in interface ContentAggregator

Parameters:

queryToContents - A map from a Query to all List<Content> retrieved with that Query. Given that each Query can be routed to multiple ContentRetrievers, the value of this map is a Collection<List<Content>> rather than a simple List<Content>.

Returns:

A list of aggregated Contents.

fuse

protected Map<Query,List<Content>> fuse(Map<Query,Collection<List<Content>>> queryToContents)