Optimize SQL query for pgvector similarity search

<dependency>
    <groupId>org.springframework.ai</groupId>
    <artifactId>spring-ai-starter-vector-store-pgvector</artifactId>
    <version>1.1.0-M1</version>
</dependency>

Currently, Spring AI uses the following SQL for pgvector similarity search:
SELECT *, embedding <=> ? AS distance FROM %s WHERE embedding <=> ? < ? %s ORDER BY distance LIMIT ?

This query calculates the <=> operator twice (in SELECT and WHERE), which may cause redundant distance calculations and unnecessary overhead. It can be optimized to compute the distance only once.

Proposed Improvement:
Use a subquery to avoid recomputing <=> and improve readability and performance:
SELECT * FROM ( SELECT *, embedding <=> ? AS distance FROM %s ) sub WHERE distance < ? %s ORDER BY distance LIMIT ?;
Do you think it'll work?

The issue provides the current and proposed pgvector similarity-search SQL but names no source file or test. Locate the Spring AI pgvector similarity-search implementation, compare the generated query and parameter handling with the proposal, and verify that results remain equivalent while avoiding redundant distance computation.