AI 문서 어시스턴트를 위한 RAG 대체 가상 파일시스템 구축

ChromaFs replaces traditional vector-search-based RAG by mapping documents to a virtual file system, enabling AI agents to navigate documentation using standard Unix commands like grep and cat. This approach reduces session initialization from 46 seconds to 100 milliseconds while eliminating infrastructure costs, proving that structured file hierarchies provide more reliable search than embedding-based similarity. The shift reveals how conventional information architecture patterns are superior for building scalable, cost-effective document AI systems.