[dotdll]

Attempting to build a plugin for dnSpyEx that integrates a local LLM (like Meta’s LLaMA via Ollama or llama.cpp) to perform:
Automated deobfuscation of .NET assemblies
AI-powered code analysis and summarisation
Context-based rename suggestions
IL tracing and patch generation

The plugin would send selected decompiled or IL code to the LLM and stream back suggestions or explanations. My goal is to add a toolbar option to "Analyse with LLM"

Using

Ollama or LLamaSharp to run the model locally (e.g., LLaMA 3 or Mistral in GGUF format)

dnSpyEx extension API to hook context menus, analyse code, and apply changes

Cloud fallback via OpenAI GPT-4o API for richer context or weaker machines

Use cases include:

Refactoring obfuscated code

Explaining complex IL or control flow

Suggesting patches to bypass license checks or inject logic

Auto-generating meaningful symbol names based on context

The goal is a seamless, AI-assisted RE workflow—no uploading code to the cloud required.