Ollama support

[digitalextremist]

Need local-only offline-first AI provider to be able to consider this IDE as an option.

Right now these are the options shown for AI provider:

Ollama support uses an OpenAI compatible API and ought to be nearly bolt-on.

[devkokooo]

Good idea, big fan of local-first models

[digitalextremist]

Exciting to see #462! Watching for a release that includes this, then going to try Athas!

[digitalextremist]

Would reopen this one @devkokooo / @mehmetozguldev if you please:

• Accepts URL for location of Ollama
  • Allows non-local IP, such as in LAN
  • Allows non-standard port
  • Allows use of Ollama Cloud

[digitalextremist]

When you get a chance, can someone please re-open this so I can check-off the test cases above?

Thanks, and much appreciated for the feature expansion to allow non-localhost URLs!

I might also be testing use of Ollama Cloud, which would require a different URL as well.

[digitalextremist]

Thanks @mehmetozguldev; keeping an eye on this one.

[mehmetozguldev]

Hey @digitalextremist! The configurable Ollama endpoint is now implemented and will be in the next release.

When you select Ollama as your provider, a new Ollama section appears in Settings > AI with an endpoint field. You can set it to any URL:

• LAN: http://192.168.1.100:11434
• Custom port: http://localhost:8080
• Ollama Cloud: just paste the cloud URL

It validates the connection in real-time and shows a status indicator. There's also a reset button to go back to the default localhost:11434.

Let us know how it works with your setup!