GitHub Copilot is nice for autocompletion, but 80% of the coding is debugging. @musabshakil shares a quick demo in his DEV post, which I encourage you to watch.

Get Instant Error Solutions with GPT

MusabShakil ・ Feb 21 ・ 3 min read

#gpt3 #ai #vscode #javascript

header image was generated using midjourney

What is MusabShakeel576/quickfix.ai?

The TLDR; is you can copy the error message and get a suggested solution from the AI. Quickfix AI is also in its alpha phase, so expect bugs and report them to the author via the GitHub repo.

quickfix.ai is an ambitious project made up of two powerful AI-powered extensions, VS Code and chrome. Both simplify your developer experience and speed up debugging.

• VS Code Extension
• Chrome Extension

How does it work?

Quickfix AI uses GPT and Vector Index to provide instant solutions for errors in your code within the code editor using AI.

The code seems straightforward, but I did not install and run the alpha build locally. I am going off the demo provided by the author in their post. I also focus on understanding the code, and the backend/src/gpt.py seems to be powered by a single python package to manage the model. Python has been the premier language to support NLP for some time, so it makes sense that the developer experience packaged through one interaction.

# /backend/src/gpt.py
from pathlib import Path
from gpt_index import GPTSimpleVectorIndex, Document

def generate_index(workspace):
    file = workspace.name+'.json'
    path = Path.cwd() / "storage" / file

    if path.is_file():
        index = GPTSimpleVectorIndex.load_from_disk(path)
    else:
        documents = [Document(document) for document in workspace.documents]
        index = GPTSimpleVectorIndex(documents)
        with open(path, 'w'):
            pass
        index.save_to_disk(path)

    return index

def generate_solution(workspace) -> str:
    index = generate_index(workspace)
    response = index.query(workspace.input)
    return response

If you understand Python and are interested in supporting it, it is open-sourced. Support the author with feedback and a star on their GitHub project.

Also, if you have a project leveraging OpenAI or similar, leave a link in the comments. I'd love to take a look and include it in my 30 days of OpenAI series.

Find more AI projects using OpenSauced

Stay saucy.