Recently I found myself using ChatGPT a lot, I was at a point where I even left Google and Stackoverflow (my two best coding friends). But ChatGPT has two problems: If I don't provide my source code as context to ChatGPT, it behaves like a large language model and provides some random answers from its brain, which mostly aren't useful and meaningful. So I've to make it pretend like "ChatGPT, you are now a software engineer" by writing some prompts curated by awesome people at . awesome-chatgpt-prompts Even if I provide my source code to ChatGPT, then write my query or code error which I want ChatGPT to solve, most of the time my source code is too big it just exceeds the . ChatGPT input limit Solution 💡 After thinking about this problem I came up with a plan to build a VS Code extension which can directly extract source code from IDE, so I don't have to copy paste the code every time I've to ask ChatGPT for the answer of errors in my code, and I use vector database as an index to store embeddings for my source code, which basically do an approximately nearest neighbors similarity search to just pick those snippets from my code which are the causing the error, in this way I'm able to input only a small chunk of source code into the GPT model which performs well and provide accurate solution for the errors without exceeding the OpenAI GPT input limit. I made the VS Code extension publicly available for free on GitHub. It's currently in Alpha stage, and I would appreciate your Pull Requests/Contributions to this open source tool. Before diving into the extension, let's understand why we need it. What are the existing solutions? 🤔 Today's Gen AI dev tools, such as Copilot, are missing the mark. Yes, they help with code completion, but every developer knows that programming is really 20% coding and 80% debugging. I asked myself, how can LLMs help with the crucial, time-sensitive, frustrating, and just incredibly intense task of debugging? Introducing 👋 Quickfix AI, an extension for VS Code that provides you within the code editor using AI. instant solutions for errors in your code With , you can: quickfix.ai 🧠 Get instant solution for error in real-time 💡 Increase your coding understanding and efficiency ⏳ Save time and minimize frustration with clear code solutions 🔍 Improve your coding skills with in-depth code analysis Demo 📽 https://www.youtube.com/watch?v=qz2U6S5tWyQ&embedable=true Installation 📦 Quickfix AI is in its Alpha phase, if you would like to use and explore it at this stage, you can follow the steps below to set it up. However, if you prefer to wait for a more reliable version with an easy installation process through VS Code Marketplace, you can for the Stable release. Join the Waitlist Prerequisite: Python v3.10 + Poetry Node v16 + Yarn Setup: Clone Quickfix AI repo Build the vscode-extension make install-vscode-extension Build the backend make install-backend Update the environment variables in both folders, (you probably don't need to update this file if your backend will be running in the localhost or 127.0.0.1) and /vscode-extension/src/env.ts /backend/.env Run the backend make run-backend How to Use Quickfix AI 🛠 Open a workspace/folder in VS Code Open the Command Palette (press or on Mac) Ctrl + Shift + P Cmd + Shift + P Type "Quickfix AI" and select the command from the list Paste/enter your error or question. Wait for the response and the solution will be displayed in a "solution-by-quickfix-ai markdown" file in your VS Code editor Enjoy the instant and comprehensive error solutions with Quickfix AI! 🎉 Also published . here
