iterative-retrieval

Multi-agent workflows encounter context errors when subagents lack necessary codebase context. Progressive context refinement iteratively dispatches broad queries, evaluates relevance, and narrows search criteria up to three times to gather sufficient context.

The best way to retrieve unpredictable codebase context is iterative search. This method dispatches broad queries, evaluates retrieved snippets, and refines search criteria based on feedback until exact files are identified.

Progressive context refinement works by iteratively dispatching broad queries, evaluating retrieved content for relevance, and refining search criteria. It loops up to three times to gather sufficient context for code-intensive tasks.

Yes, iterative search is ideal for complex bug fixes. It intelligently searches for relevant files like tokens and sessions, progressively refining search criteria until the exact files needed for the fix are identified.

Dynamic context gathering is needed when subagents cannot predict required context upfront, such as during complex bug fixes or new feature implementations. It prevents missing context errors by evaluating and refining search results.

Iterative context retrieval loops up to three times to gather sufficient context. If the codebase context is too large or the search criteria cannot be refined effectively within these iterations, the retrieval may not find all relevant files.