Anyone run a local LLM against a code base to give it context and ask questions about it?

Reply to this note

Please Login to reply.

Discussion

What's local LLM code?

https://github.com/bublint/ue5-llama-lora

Large Language Model. But I want to “train” it specifically on my code bases without having to send code off to someone else servers.

Did you try gpt4all?

Yup! But it lacks current contextual knowledge of projects being written today.

Example I may tru to emulate: https://github.com/bublint/ue5-llama-lora

No but I discovered PDF.AI which is really useful and a privacy nightmare rolled into one.

👀

Not a code base but been doing it with PDFs - it wouldn’t be difficult to adapt a solution meant for text to work with code. Look into using a vector database (like Pinecone) to store embeddings and you’ll be on the right rabbit hole path

My inspiration came from this:

https://github.com/bublint/ue5-llama-lora

I want to replicate this but using a concatenation of an entire codebase + docs.