chore: Convert actions to Kotlin

carlrobertoh / CodeGPT

The leading open-source AI copilot for JetBrains. Connect to any model in any environment, and customize your coding experience in any way you like.

https://codegpt.ee

Apache License 2.0

1.06k stars 223 forks source link

chore: Convert actions to Kotlin #482

Open reneleonhardt opened 6 months ago

reneleonhardt commented 6 months ago

I suggest streamlining all actionPerformed:

Send Telemetry if nothing happened?
Show dialog if action is not possible?
Throw NPE with requireNonNull() or handle null accordingly?

reneleonhardt commented 6 months ago

@carlrobertoh I converted conversations on top of this branch, any chance that you can merge this first? Another 29 files changed, still 120 to go 😅

carlrobertoh commented 6 months ago

I haven't had a chance to review this PR yet, but I'll try to do it sometime early next week. Also, I'll probably do the next release without these refactorings because our test coverage is bad, and I don't want to risk breaking anything.

reneleonhardt commented 6 months ago

Sounds good 👍

reneleonhardt commented 6 months ago

Speaking of which, please try Llama 3 locally first before releasing 😅

carlrobertoh commented 6 months ago

The response seems a bit off. Also, the success callback isn't being triggered once the server is up and running, so it looks like it's booting up forever.

carlrobertoh commented 6 months ago

Was able to fix the response by modifying the prompt and adding a stop token.

reneleonhardt commented 6 months ago

Thank you for testing and fixing! Stream was not the problem, begin_of_text isn't needed, but assistant, stopTokens and server hack (?). Are other models running with msg instead of message, is the hack "official"?

carlrobertoh commented 6 months ago

Thank you for testing and fixing! Stream was not the problem, begin_of_text isn't needed, but assistant, stopTokens and server hack (?).

I followed the template format from their official guide. Even if it's not required, I'd still like to have it included unless it breaks something.

Are other models running with msg instead of message, is the hack "official"?

I called it a hack because the way we detect if the server is up and running is through the server logs. I assume they changed their logging structure (from message to msg), and that broke our trigger condition.

carlrobertoh commented 6 months ago

Hey, I'm putting this refactoring on pause for a bit until I have run a compatibility comparison between different IDE builds. Most likely, this won't cause any issues, but I just want to be sure, since I'm slowly considering supporting older versions again.