Show GN: turbo-graph – turbovec에 그래프 메모리/필터 캐시를 얹은 constrained RAG 인덱스

turbo-graph addresses a critical pain point in modern RAG systems: the need to apply multiple overlapping constraints (ACLs, metadata filters, temporal windows, graph relationships) efficiently. By moving graph memory, metadata compilation, caching, and reranking into the index layer itself, it reduces engineering overhead that typically spans across Python, SQL, and application layers. This enables simpler RAG pipelines that integrate vector search, graph traversal, and metadata filtering as a unified operation.