iterative-retrieval

Iterative retrieval prevents subagent context overload by dispatching targeted queries and evaluating file relevance across up to 3 cycles. Instead of sending entire codebases, it progressively refines search criteria to collect only high-value context.

Iterative retrieval is a 4-phase dispatch-evaluate-refine-loop workflow that solves subagent context limitations. It uses relevance scoring and gap identification to progressively gather targeted context for tasks where initial requirements are unknown upfront.

You can collect relevant code context by applying iterative retrieval to progressively refine search terms based on discovered codebase terminology. It evaluates file relevance and identifies context gaps across 3 cycles to eliminate guesswork.

Use progressive context refinement for code exploration, bug fixes, or feature implementation when initial context requirements are unknown. It adapts search patterns based on discovered terminology to gather targeted context and prevent missing information failures.

Yes, iterative retrieval applies to RAG pipeline design by using a 4-phase loop with relevance scoring and gap identification. It progressively collects targeted context across 3 retrieval cycles instead of sending full data upfront.

Iterative retrieval limits agent orchestration to 3 retrieval cycles for collecting targeted context. It requires evaluating file relevance and refining search criteria at each phase, making it unsuitable for tasks needing full codebase delivery upfront.