You are right. We do send it to OpenAI, and it's mentioned in a couple of lines below. I meant to update the readme to mention it everywhere but forgot! I just pushed a fix.
This is still far better than uploading your codebase to a third-party service for indexing and embedding generation. This tool intends to solve that right now with OpenAI's function calling.
However, the dream goal is to eventually run an LLM locally on your machines and support search without talking to any third-party service - https://gpt4all.io/index.html. But they are yet to arrive; none of the open-source LLMs are even close to the level of performance or quality as GPT 4 yet.
This is still far better than uploading your codebase to a third-party service for indexing and embedding generation. This tool intends to solve that right now with OpenAI's function calling.
However, the dream goal is to eventually run an LLM locally on your machines and support search without talking to any third-party service - https://gpt4all.io/index.html. But they are yet to arrive; none of the open-source LLMs are even close to the level of performance or quality as GPT 4 yet.