hierarchical-indexing

TreeRAG: Hierarchical RAG platform that indexes dense PDFs into logical trees for precise retrieval, 100% traceability, and LLM-guided deep traversal. Production-ready with enhanced security and streaming indexing.

Dynamic Segmented Memory (DSM): A hierarchical graph-based retrieval engine for scalable LLM context management and associative reasoning.

Vectorless, reasoning-based RAG using hierarchical document indexing with Vertex AI. No embeddings, no vector DB - just LLM reasoning.

Fast entity tree RAG with Cuckoo Filter acceleration. Hierarchical entity indexing cuts retrieval latency up to 85% while preserving accuracy.

把检索系统的「方向」做成可插拔的分层结构：一层比一层细，每一层只在下层表达不出东西时才建。方向由人声明，方向内部的参数交给使用 —— 没有全局目标函数，不排序、不打分、不做匹配。 | A pluggable layered direction view for retrieval systems: each layer finer than the last, built only when it says what the layer below cannot. Directions are human-declared; no global objective, no ranking, no scoring.