Chapter 5. Considerations Using AI with Enterprise Codebases
We recognize that we are all in a world of experimentation with AI. Many development teams are gravitating toward using AI assistants or chatbots to help generate code. Even with the hallucinations, this is an important application of AI for code.
However, AI assistants are not built for mass-scale code changes—many individual developers with their AI assistants operating on one file, one cursor position at a time. This by definition limits the problem space and requires each developer to review the output. To apply AI to maintaining and securing large-scale, proprietary codebases, you need some different approaches as we’ve shared in this book, with the end goal of making accurate changes to codebases at scale.
LLMs need to be augmented with the code data via various techniques. RAG is one that’s popular today, but others may appear in the future. In addition, instead of embedding code as text like any other business document, you can do one better if you take into consideration the structured nature of the code. When you understand the code and are able to extract relevant methods or classes from code, the better you can guide the model with just enough relevant information. These approaches are useful whether you decide to use LLMs via API or plan to deploy and operationalize them from OSS.
As you take your next steps with AI in your software development organizations, we want to leave you with our top considerations ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access