autopilot VS ChatGPT Auto Refresh

Thanks for sharing your work! Let’s connect cause it seems we are exploring a similar space. Here’s the project I have been working on: https://github.com/fjrdomingues/autopilot

- https://github.com/fjrdomingues/autopilot/issues/178 (the code is mostly correct, minus a minor change in the package that was updated and gpt doesn't know. It was almost a copy-paste).

Code Autopilot is an AI-powered tool designed to tackle software engineering tasks. It shares similarities with Github Copilot, but with some key distinctions: - It derives context from your entire codebase, enabling it to handle complex tasks spanning multiple files. - It integrates with your Github account to obtain context and respond to issues you open. - The core technology is open-source, from https://github.com/fjrdomingues/autopilot

Consider using autopilot for coding https://github.com/fjrdomingues/autopilot It’s open source In my tests it works with gpt-3.5 but gpt-4 gives much better/reliable answers. Problem is the price difference

The solution I found was to create an agent that summarises all the files in a project and then a sequence of agents that picks the necessary code/folders for a task. That way I was able to provide the required context and then let GPT do its magic. The project is public at: https://github.com/fjrdomingues/autopilot

Check what we are doing in https://github.com/fjrdomingues/autopilot we have multiple use cases were we ask gpt 3.5 to reply in JSON

Link to project: https://github.com/fjrdomingues/autopilot

I developed a too for the same problem but with a slightly different approach: https://github.com/fjrdomingues/autopilot It uses multiple GPT calls to get context on the codebase and when you give it a task it uses the context to understand where to implement the code changes and suggests the code to implement. It would be great to merge the best of both project