Need help relating to Neo3j cypher

Graph Academy & Newbie Questions Newbie Questions
1

hi i’m neo4j beginer.
i’m curious if there’s python library or package about what i wanna make.
as i have speech on next wednesday, i need help from people here
i’ve tried to get some evidences from youtube and some articles from internet.
but i and ChatGPT fails…. please help
i have ChatGPT, Cursor, Neo4j desktop 2.0.5. Docker Desktop. Visual Studio Codes. python 3.13.3

what i wanna make is something like this.

  1. scan documents (docx, txt, md file format) from one folder.
    let’s say there’s only one file inside.
  2. divide it by some paragraph headers level like (1. , 1.1, 1.1.1). hierarchy. and all headers becomes nodes(vertex, entity) with properties.
  3. when making hierarcy, i want LLM to make edges(relationships) and name it by itself. so i don’t need to make all the relationships between basic nodes.
  4. embedding those Data in Graph DB

is there python package or library any like what i say?

and thing is, i wanna have some real check if steps are going well, not making mistakes like, wrong headers becoming nodes. but i don’t know to check it and fix it yet.
i have tried to make sample DB and i couldn’t figure if my DB is making edges well. and don’t know how to fix it

sorry for hard question.
ah if you are Korean native, please write it with Korean language please.
i wish you have good day

2

Hi,

There is a demo application where you can “try out” with a couple of your documents here: https://llm-graph-builder.neo4jlabs.com/ I think it may be good to start there and run a couple of iterations so you can figure out what entities and relationships you want to focus on extracting.

Under the hood this application is using this python package GraphRAG for Python — neo4j-graphrag-python documentation

But there are many more frameworks you can use. If your documents have a “known structure” you may want to write some custom code for the “chunking part”.

To extract entities and relationships, you can define your domain model in pydantic and use the openai package for instance (small trivial example here: notebooks/Module_02_GraphSummit_Nordics_2025.ipynb at main · lqst/notebooks · GitHub )

Good luck!

3

Thank you for your kindness : )