search

getcursor / cursor

The AI Code Editor
https://cursor.com
20.63k stars 1.38k forks source link

Please allow us to specify token limits when using our own API key #1370

Closed grass29 closed 3 months ago

grass29 commented 3 months ago

Is your feature request related to a problem? Please describe. Currently the token limit is set globally, which severely limits the potential when using our own API key.

Describe the solution you'd like Allow us to specify custom limits, ideally per model and output/input tokens (so Claude would be 200k, Gpt-4-turbo would be 128k).

Additional context The user should have more control when using their own API key, this would significantly help users who need to tackle a difficult code problem where huge context is important, and don't mind spending extra doing that.

Dreamzlol commented 3 months ago

As I mentioned a couple of weeks ago in this thread: https://forum.cursor.sh/t/capped-at-10k-context-no-matter-api-key-or-cursor-pro/3442 Also using docs with just a 10k window is pointless.

Please let us choose how much we limit the context. Sometimes, I have more than 3-4 files that I need to add for context. And i am not the only one it seems. Would be great, Cursor Team!

grass29 commented 3 months ago

Seems to be possible now: https://forum.cursor.sh/t/solved-please-add-claude-3-opus-200k-and-gpt-4-turbo-128k-and-gpt-4-vision-128k-to-long-context-chat-when-on-our-own-api-key/3948?u=fun_strange