memory-retrieve

To avoid overwhelming AI context windows, use a progressive disclosure approach to retrieve minimal relevant memory. This mechanism scans an index file and selectively loads individual ADR files based on domain and level relevance, preventing context contamination.

Progressive disclosure for AI memory retrieval is a 3-level system that fetches memory incrementally. It applies to AI agents requiring historical context by starting with an Index Scan, moving to Individual ADR Files, and finishing with a Cross-Domain Scan for decision-making and planning.

To selectively load architectural decisions and patterns for a specific feature domain, activate a memory retrieval mechanism that fetches only necessary durable memory. This ensures relevant historical context is loaded for the specific domain rather than dumping the entire project's memory.

Yes, progressive disclosure works for AI agents requiring historical context across different domains by implementing a Cross-Domain Scan as its third retrieval level. This fetches relevant decisions and patterns incrementally for planning and implementation phases without full memory dumps.

You should avoid loading the entire project memory into your AI agent's context when implementing features or making decisions. Loading full memory dumps risks context contamination and reduces efficiency, making selective loading of durable memory based on domain relevance the preferred approach.