Analyzing a source code repository, uploading its representation to a neo4j graph database, querying the db to validate code references and detect hallucinations is next level.
Discussion
How we can turn ANY code to a graph data set?
I’m using this right now:
Wow! It already helps me for a lot of stuff I'm planning.
What you are using this for?
To check what has already been done, but I also keep a TODO.md with the items that need to be done for each phase, and a CHANGELOG.md the LLM can always read.
To check for hallucinations: is the new code passing an argument for a parameter that doesn’t exist or has a different type, or calling a function that isn’t there.
The next step for me is creating graphs for dependencies, so the LLM can detect hallucinations on them.