iterative-retrieval

To find relevant files and context for complex debugging, you can use an iterative search approach that runs up to three cycles of dispatching queries, evaluating relevance scores, and refining keywords to surface pertinent code across multi-repo projects.

Iterative code discovery is a multi-cycle search process that evaluates and refines retrieved sources. Relevance scoring works by assigning a 0.0 to 1.0 priority value to candidates, helping you filter highly pertinent code and documentation from less relevant results.

To locate the right files for large-scale refactors, apply a looped refinement process that re-evaluates initial candidates and re-searches with new keywords, continuously refining context collection until the correct background knowledge is surfaced.

You can automate background knowledge collection for feature development in multi-repo projects by initiating a structured search cycle that dispatches queries, evaluates relevance, and loops through refinements to clarify complex task contexts.

The best way to surface relevant code when a single search is insufficient is an iterative retrieval method, which executes up to three search phases that evaluate and refine results based on dynamic relevance scoring and contextual gaps.

Iterative file discovery works for deep context tasks by applying up to three search cycles across multi-repo projects, utilizing dispatch, evaluate, refine, and loop phases to thoroughly clarify complex task requirements and relevant background knowledge.