iterative-retrieval

To retrieve the right codebase context, use an iterative retrieval loop that progressively refines file search criteria. This method applies a DISPATCH, EVALUATE, REFINE, and LOOP workflow to narrow broad candidate files down to the most relevant context for your subagent.

Prevent missing context failures by applying an iterative retrieval loop with relevance scoring and gap detection. This approach progressively narrows file candidates and identifies missing critical information until enough high-relevance context is retrieved.

Optimize token usage during file retrieval by implementing a maximum of three refinement cycles with an early-stop rule. This strategy limits retrieval loops and stops early once sufficient high-relevance context is identified for the debugging pipeline.

Relevance scoring assigns a score per retrieved file and identifies critical information gaps during the EVALUATE phase. The query is then refined to target those gaps, progressively improving target relevance for your feature implementation scenarios.

Yes, you can use iterative retrieval for multi-agent workflows with unpredictable context requirements. It solves the subagent context problem by progressively refining file retrieval criteria through up to three cycles of query refinement and missing-context gap identification.