I've got qwen-coder, deepseek, llama and a few others running on Ollama and connected to IntelliJ via the Continue plugin, but none of them are going to give my anything useful here. If you look at the minimal example, it will work on just 5~10 files. The issue is with 30~50+ files. Somehow the view isn't ready to have an index set, even though I'm ensuring the model is fully refreshed.

An LLM isn't going to be able to help you find some weird race condition. They're best for simple, know problems or for quick documentation reference.