This repository was archived by the owner on Apr 1, 2025. It is now read-only.
DependencyAwareRefactoringService requires fulltext and/or hybrid search capabilities #7
Description
The current design of DependencyAwareRefactoringService returns too many refactor candidates resulting in occurrences of HTTP/500 - org.springframework.ai.retry.TransientAiException.
We need a way to refine and constrain similarity search constraints to honor full text within documents in addition to metadata.
The Spring AI team is working thru some features, tracked here:
- Add 'like' operator in FilterExpressionBuilder spring-projects/spring-ai#328 - like searches on metadata attributes
*Enhanced Vector Store Capabilities with Full-Text/Hybrid Search and Reranking spring-projects/spring-ai#1227 - full text and hybrid search
Once these features become available, we will need to incorporate them into the current implementation.
Prompt engineering for refactoring is also worth some further investigation.