The AI Agent needs to be provider agnostic

[IAmTomShaw]

The use of GitHub Copilot SDK is a fantastic feature in this project. However, it potentially restricts other users who don't have a copilot subscription from using the project.

Adding a menu which allows the user to select their chosen provider and set their credentials would be a great solution for this and make the tool much more accessible to everyone in the community.

[ArturCharylo]

Hey @IAmTomShaw!

I've been analyzing the issue regarding multi-provider support and I'd love to take a crack at it. Since the project currently relies heavily on the Copilot SDK, I'm proposing a more modular architecture to make the tool accessible to everyone.

Here is my proposed plan:

1. Provider Abstraction: Create an AIProvider interface to decouple the core logic from specific SDKs. This will allow us to easily swap between Copilot, OpenAI, and potentially others (like Anthropic or local Ollama).
2. Configuration Management: Implement local storage (e.g., using the conf package) to persist the user's provider choice and API credentials securely.
3. Setup Wizard: Add a first-run experience that guides the user through selecting their preferred provider and entering their keys.
4. Modern Model Support: For the OpenAI implementation, I'll ensure we support the latest models like gpt-5.2, while perhaps allowing users to specify a custom model string in settings.

Do you have any specific preferences regarding:

• How the configuration should be stored (JSON in config dir vs. .env)?
• Whether we should support a "fallback" chain if one provider fails?

Let me know if this direction aligns with your vision for the project!

[IAmTomShaw]

This sounds great!! Is this something that you'd like to build? :)

[ArturCharylo]

Absolutely! I've actually just finished refactoring the core logic into the AIProvider interface locally. I'll start working on the OpenAI implementation and the CLI selection menu next. I'll open a Draft PR soon so you can see the progress!