CNodeChat

LLM RAG chat for VS code for C/C++ projects. It parses your project with libclang, save function calls into database and LLM Agent could enlarge its context with related code.

Modules

• exstension - source files for vs code exstension
• service - python backend, that act as http server and interact with ollama service AND script for build faiss database.

Demo

Diagram

1.                                         ---source_codes--->[ProjectParser]--VectorDB--->[RAG]

2. [VS Code extension] -user_query-> [Uvicorn server] -user_query-> [Ollama] -llm_query--------> [RAG Agent] --query_to_db--> [RAG]
                                                                            (optional tool call)          (mandatory tool call)

3. [VS Code extension] <-answer------[Uvicorn server] <-answer----- [Ollama] <-filtered_answer-- [RAG Agent] <-long_answer--- [RAG]

How build:

• install conda enviroment from enviroment.yml
• start ollama service:

ollama serve

• build database with create_database.py, don't forget to choose dir
• start oll_chat.py
• open air_chat_exstension/ with vs code, then start it

TODO

• Refactoring
• clang AST based context enlargement
• add database build tools in frontend
• more refactoring
• fix frontend
• containerization for service
• distribution for exstension