Knowledge graph is quite big

[bhupesh-sf]

Hey, I tried your tool and its quite good. It works very fast. Only challenge I am facing is knowledge graph for a medium size project also quite big and hence not able to send it to LLM. In my case its around 200000 token

[JohnRodney]

@bhupesh-sf I've found it really helpful to just put the info into a neo4j db running in docker and let the AI query it. Give it a few analysis tools on an MCP and you get quite a lot of information about your codebase fast without having to provide unimportant stuff. I ended up adding all the class and fn context to the parser so I could store the code itself and the AI is able to tell me about duplicate code patterns across the codebase and design refactors that are pretty good.