Show HN: Open-Source Alternative for GitHub Copilot

[suneelak]

https://marketplace.visualstudio.com/items?itemName=flexpilot.flexpilot-vscode-extension

After many months of effort, trying to build an open source alternative for GitHub Copilot, while still retaining the same native user experience it offers, we're happy to ship Flexpilot. (https://flexpilot.ai/)

You can try the VS Code extension here - https://marketplace.visualstudio.com/items?itemName=flexpilo...

Flexpilot requires VS Code 1.95 or higher.

While we loved GitHub copilot and the experience it offered, we wanted to use other LLM's to code effectively using VS Code and not just OpenAI. Also, there are lots of concerns around privacy and security at enterprises to send their codebases to the cloud. Several other coding assistants did not have the same native experience as GitHub Copilot and were only web views and we wanted to try and build something that will give us the best of both worlds - flexibility to choose LLM's (Cloud or on-prem) and native user experience, all within VS Code.

Flexpilot offers you the ability to choose from a variety of LLM providers such as OpenAI, Claude, Mistral, Gemini, AzureOpenAI, Groq or even other OpenAI compatible API providers. For instance, you can run Llama 3.2 or CodeLlama using Ollama on your local machine and inference it on VS Code using Flexpilot.

Not only in panel chat or inline chat, Flexpilot allows you to choose any LLM of choice in Panel, Inline, Code completions, quick chat, commit messages, Chat summaries etc.

Our repo is open source and you can find the details here - https://github.com/flexpilot-ai/vscode-extension

Happy to answer any questions on tech stack, comparisons with other coding assistants, etc.

Please try Flexpilot and let us know your thoughts!

[thedragula]

What's your strategy for maintaining the "native" VS Code experience across different LLM providers, given that they might return responses in different formats? How do you standardize the interface while preserving each model's unique capabilities?

[suneelak → thedragula]

Hey, thanks for asking. The LLM's do not hamper the native experience in anyway, as the UI/UX is controlled by the Flexpilot plugin and not the LLM. The interface is designed to be standard across all LLM's. The accuracy of the responses may vary according to LLM's but that is a different topic for evals.

[sree2024]

Very friendly for beginners.

[suneelak → sree2024]

Thanks! Is there any new feature you would want to see in an AI Coding assistant? If so, would you be able to pass on some details?