Tips on speeding up Cursor?

How To
1

I’m a beginner using streamlit, a python-only web framework, and so far I prefer to throw everything into one giant script (so far 4000-5000 rows) rather than separate files per page or function.

This doesn’t seem to gel well with how Cursor works though…
with cmd+L chat assistance, it seems to be reading the whole file every time (as well as the rows I’ve selected before cmd+L).

When I want to apply, it strangely adds the edit quickly then gets stuck ‘generating’ and usually times out.

So is it worth distributing my functions into separate files perhaps?

– I am using cursor pro with plenty of fast queries in the bank.

2

Out of interest, have you:

  • Enabled Long Context Chat in Cursor Settings > Beta

  • Set the ‘mode’ of the Ctrl + L chat to Long Context Chat
    (see the mode dropdown in the top-right)

  • Chosen a Long Context Model, ie:

    • gpt-4o-128k
    • gemini-1.5-flash-500k
    • claude-3-haiku-200k
    • claude-3.5-sonnet-200k

And a few other troubleshooting questions:

Are you using any of the @ symbol features?

Which button are you choosing in the chat interface:

  • Chat with context
  • Chat with no context
  • Chat with codebase

Which LLM were you using?

Can you provide system information from Cursor > Help > About?