HamedMP/CursorLens

Failing to update API endpoint in Cursor Settings

Closed this issue · 4 comments

Everything works fine except error while verifying the endpoint in Cursor Settings.

It just doesn't work, but the curl version show by Cursor does work.
Screenshot 2024-08-19 at 8 53 15 AM

Anything I miss out?

Hi,

No need to press Verify as the model is set on the Configuration page and the api key is the one in ENV variable.

You only need to override the OpenAI base URL and choose any gpt model in Cursor so the OpenAI URL is chosen.

final checklist:

  • the base url should be updated
  • the toggle should be on
  • The key is not important - and verify wouldn’t work

You can test with: cmd+L and send a no context chat to a gpt model, it should be routed through CursorLens.

Let me know if you have more questions.

Does the API endpoint have to be an NGrok URL, or can I also use localhost?
The latter would allow using Cursor with local models even when no Internet connection is available.

Thanks for the question @nightscape , Cursor requires https models, so the current workaround is ngrok or similar service, but we're looking into ways to do self-signing certificates to run localhost with https to fix this issue and be completely offline/off the grid 🚀

Thanks @HamedMP 👍. Just tried again and it works!
However, the toggle button only show if you have successful verify before, so you somehow still have to click Verify once with a valid Key then only change the base URL.