easy sir, stop abusing. this is a just a trivial graphrag idea. 1. you can use llm to automate this process, refer to https://github.com/microsoft/g... GraphDB, like neo4j, or nebula. 3. can be embedded as vector, do vector similarity search on nodes. omg.