OpenSearch:Develop sort plugins using Cava

Step 1: Create a sort policy

1. Go to Policy Management and click Create.

2. On the Create Policy page, set Scope to Fine Sort and Type to Cava Script.

Step 2: Add script files

Add script files using either method:

• Edit the Cava script directly in the console editor.

• Upload a local script file in JSON format.

Step 3: Compile and publish

1. Click Compile to compile all script files. The compilation status appears below the Compile button.

2. After compilation succeeds, click Publish.

Step 4: Test the sort policy

On the Search Test page of the policy, set the following parameters:

The following Java example shows how to set these parameters in the SDK:

// Create a SearchParams object
SearchParams searchParams = new SearchParams(config);

// Configure the Cava sort policy
Rank rank = new Rank();
rank.setSecondRankName("Cava script name");    // Exact name of the published sort policy
rank.setSecondRankType(RankType.CAVA_SCRIPT);  // Invokes Cava script during fine sort

// Attach the rank config to the search request
searchParams.setRank(rank);

Vector index score

Use ProximaScore to retrieve the similarity score from a vector index. Call ProximaScore.create() to bind to a named index, then call evaluate() to get the score for each document.

import com.aliyun.opensearch.cava.features.similarity.ProximaScore;

// Bind to the target vector index
ProximaScore _proximaScoreVector;
_proximaScoreVector = ProximaScore.create(params, "{{indexes.vector_index}}"); // Replace with your vector index name

// Get the similarity score for the current document
float proximaScoreVector = _proximaScoreVector.evaluate(params);