Company
Date Published
Author
Michael Hunger
Word count
2007
Language
English
Hacker News points
None

Summary

The LLM Knowledge Graph Builder is an open-source tool that enables developers to build retrieval-augmented generation (RAG) experiences to interact with unstructured data. It preprocesses documents, transcripts, and web articles into chunks, computes text embeddings, and connects them to create a lexical graph. The tool also extracts entities and their relationships, which can be used to connect pieces of information and provide relevant context to user questions. In the first release of 2025, the LLM Knowledge Graph Builder introduces new features such as generating community summaries, running multiple retrievers in parallel for question answering, guiding extraction with custom prompt instructions, and supporting read-only database connections. The tool has been tested with various large language models (LLMs) and has shown promising results. It aims to improve the richness of knowledge graphs representing unstructured data and provide a more interactive experience for users.