CodeAid

• 

  Search for github repositories

• 

  Ask questions about how to work with the code in the repository and ChatGPT will load in context

Inspiration

We like using ChatGPT to help us design code or write scripts, but making sure it has enough context to give a valuable answer usually requires copy/pasting descriptions of your service and how it's structured. CodeAid automates and improves on this by allowing ChatGPT to look up any repo on github and provide the most relevant parts of its documentation as context for its answer.

What it does

CodeAid indexes python code and markdown files using embeddings and provides ChatGPT with an API to search for indexed content.

How we built it

We built a python flask server that uses Docarray and annlite to create and store indexes of embeddings.

Challenges we ran into

We wanted to generate summaries of the python code in code bases, but found that generating summaries took too long to do effectively.

Accomplishments that we're proud of

The tool is powerful to use and even in its simple state makes chatGPT significantly more useful to help with coding problems

What we learned

Integrating an API with ChatGPT is extremely simple, small simple plugins like this should be spun up for most services that contain useful data for working with ChatGPT.

What's next for CodeAid

We would love to improve the scalability of our index and improve how we index code files to make CodeAid even more powerful for code bases that aren't fully documented.

Built With

• docarray
• flask
• python