Show HN: Knowhere-Rebuilds the document structure that chunking throws away

Category: ai-ml

Tags: rag, document-parsing, ai-agents, knowledge-graph, llm, retrieval

Score: 7.5/10 (Innovation: 7, Technical: 8, Documentation: 8, Utility: 7)

Knowhere is a document parsing and retrieval system that reconstructs document hierarchy and structure (trees, sections, cross-document graphs) instead of flat chunking, then exposes this as 'navigable memory' for AI agents. It combines a dual Vision/Text parsing track with an agent-native retrieval API, positioning itself as memory infrastructure for Agentic RAG rather than another PDF extractor.

Target audience: AI/ML engineers, RAG developers, backend engineers building agent systems

Repository: https://github.com/Ontos-AI/knowhere · Python · Apache-2.0 · 3330 stars

View on Hacker News