Docs: diff two Microsoft GraphRAG runs

Open Beginner friendly
  • Is something wrong?
#18 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
2/5
Estimated time
1-3 hours
Newbie friendliness
78/100
Issue type
Documentation
Clarity
Clearly specified
Activity status
Active
Tech stack
pandas, python, typescript
Domain
documentation

Research direction

Read docs/guide/llm-graphs.md around the Property graph formats section and check GraphRAG's entity and relationship Parquet column names for the version you document. Add the pandas conversion snippet, the comparison command, and a brief explanation of the results; update docs/zh/guide/llm-graphs.md if possible. Done when the English section is clear and the Chinese version is added or its omission is noted in the pull request.

Written by the indexing model from the issue text.

Description

documentation good first issue
Why

Microsoft GraphRAG writes its knowledge graph as
Parquet tables (entities and relationships), and its ids change on every run. Comparing two
runs, after changing a prompt, a model or the input documents, is exactly what kgdiff's
similarity matching is for
(Graphs without stable ids). What is
missing is the few lines of Python that turn the Parquet tables into a file kgdiff reads.

What

A short section in docs/guide/llm-graphs.md, after
Property graph formats,
with:

  1. A Python snippet (pandas only) that reads the entities and relationships tables of one run
    and writes kgdiff JSON: the
    entity title as label, its type as type, the description and other columns as
    attrs, and relationships as edges. Check the column names against the GraphRAG version
    you use and say which version the snippet was tested with.
  2. The command to compare two runs, e.g.
    kgdiff run-a.json run-b.json --match similarity --ignore-attr <columns that change every run>.
  3. One or two sentences on what to look at in the result.

The same section in Chinese in docs/zh/guide/llm-graphs.md if you can, otherwise say so in
the pull request and someone will add it.

How to check it

pnpm dev:docs serves the docs locally. If you have two GraphRAG runs at hand, run the snippet
on them and mention in the pull request what kgdiff reported.

This is a good first issue: no TypeScript, one page, and a real gap for GraphRAG users.

Dominant language
TypeScript
Stars
47
Forks
6
Avg merge
2h 32m
Merged PRs (30d)
22

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from iiwish/kgdiff

All issues in iiwish/kgdiff

Similar issues

More TypeScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.