Closed Dawn-Xu-helloworld closed 1 month ago
@Dawn-Xu-helloworld There are a couple of options to achieve this that you can already do within Continue. The first is that you can set a trigger key for inline suggestions.
To do this, add the following to your VS Code settings:
{
"editor.inlineSuggest.enabled": false
}
And add this to your keybindings settings file (you can choose any shortcut you'd like):
{
"key": "alt+\\",
"command": "editor.action.inlineSuggest.trigger"
}
The second option is to increase the debounceDelay in Continue's config.json so that it won't send a request on every keystroke: https://docs.continue.dev/walkthroughs/tab-autocomplete#tabautocompleteoptions
Validations
Problem
For a no-GPU, self-host LLM, the previously compilete mode would make a lot of requests to server, but server can't deal with such a lot of requests.
Solution
So I think if we add a "request a response after keyboard idle for some times" or "after a hotkey" or after something, in order to make less requests and cost less computing resources, for a self-host server