Code Explanation / Refractoring with GitLab Data

Problem to solve

In addition to open source dataset, we would also need to understand a high quality dataset for early users and extend prompt library

1. This is for code explanation and refractoring
2. Gitlab Specefic Code Generation

Proposal

1. Looking at the opensource repos we had we would first build the templates for code explanation
2. This is followed by reverse engineering for code generation.

Further details

We would be using GPT-4 and GPT-4 turbo as evaluator model and prompt generator models in addition to claude and text-bison

Links / references