iterative-retrieval

Iterative context retrieval works by employing a four-phase loop: Dispatch, Evaluate, Refine, and Loop. It dynamically adjusts search queries based on relevance scoring and identified information gaps to progressively fetch precise sub-agent context.

Refine codebase navigation by dispatching initial search queries, evaluating relevance to identify gaps, and dynamically adjusting search patterns. This progressively pinpoints exact files like authentication modules needed for debugging without overwhelming agents with irrelevant code.

Yes, dynamic query adjustment adapts search patterns and keywords based on evaluated relevance and identified gaps. This handles insufficient initial context by progressively refining the search to ensure agents retrieve precise information for complex code analysis tasks.

The best way to prevent exceeding context limits is using iterative context refinement. It evaluates search relevance and dynamically adjusts queries to solve the context problem, ensuring sub-agents receive only necessary information without irrelevant code overload.

You need phased retrieval when initial context is insufficient for complex multi-agent tasks. It uses a Dispatch, Evaluate, Refine, and Loop cycle to progressively filter codebase information, ensuring sub-agents get precise data required for their specific task.