Companies that have sufficient history or sufficient number of engineers will have a lot of internal source code, documentation, bug tickets, CRMs or other knowledge bases that are daunting for new engineers to pick up. Even old engineers have a hard time understanding the domain space that is not his expertise. The knowledge is scattered everywhere.

So I’m looking for an “expert system” that I can install on my premise ( or have access to my GitHub/Bitbucket hosted instance), that would allow me and my staffs to query our internal knowledge system ALA Q&A style. The main idea is for new engineers to get to up speed with our code base. Essentially it is just a generic baseline model (open source LLM such as DeepSeek or LLAMA) with an internal knowledge base on top of it. The baseline model is just to give the “background” knowledge to the expert system. My internal knowledge base is the cream that will pump out useful information for me.

The system doesn’t need to be 100% accurate; my engineers just need a sounding board, or to know where to even start to read, in order to understand the code. I know it’s good also for new engineers to “struggle” with the codebase from time to time in order to generate an intuition for the whole thing. But still, for some tasks, that wouldn’t be necessary; all the engineers need is to quickly get into the right part of the source code, make changes and then get out.

How should I start if I want to write such a system? Or is there already such an out-of-the-box solution out there? If not, it would be an interesting idea for people to work on.