Thats the first success i have had in days with interpreter mode, thanks. However, it times out on longer files being edited which is a complete blocker. And @andrewh instead of just outputting the file name and a description of what its going to do, it tries to output the entire file change in the place it used to just have the file name (see below)
Can you look into this? for small files it works alright but takes way longer because your waiting for it to output the entire change first before executing. Something is different from what it used to be. Used to just have the filename to edit and a description about what its plan was
working now!
The code interpreter mode is really great at bumping versions/updating changelog/drafting commit message/doing the actual commit automatically.
I was still having some issues but after testing for a few hours if all fails and you can’t get it to work. My eventual fix: I pressed the “Copilot++” on the bottom left of the cursor IDE window and disabled the “Copilot++” globally and it the Interpreter mode began to work.
Was a bit miffed that you lot broke interpreter, and clearly don’t have metrics on what features your userbase uses.
Interpreter is IDEAL for large file edits, rewrites, refactoring and CLI stuff.
GPT-4o is WAY too verbose and outputs the entire file change instead of ide.edit(filepath/filename), “my instruction”
…
When they get it fixed, could someone post a demo of using code interpreter in an efficient way? These replies got me really curious and make me wonder if I’m using cursor to its fullest potential. Also, probably being snarky here, but cursor really needs to do some user interviews because it’s clear that their schema for code editing and use is a bit different than some of their paying users.
Interpreter Mode has returned to work without problems in version 0.37.1. Prompts containing multiple tasks can now be processed autonomously at once with a single ‘auto-excute’ click.
?