When I tried to use “@” in Cursor to load my code file, it seems Cursor couldn’t read the file.
Even after showing the code directly, Cursor kept asking me to send the code file. I’m not sure why, but this issue seems more likely to occur after prolonged usage or extended conversations with Cursor.
Here are the details of my Cursor version:
• Version: 0.44.11
• Commit: 1.93.1
• Date: fe574d0820377383143b2ea26aa6ae28b3425220
• Electron Build ID: 30.5.1
• Electron: 2025-01-03T07:59:06.361Z (1 week ago)
• Chromium: Undefined
• Node.js: 124.0.6367.243
• V8: 20.16.0
• OS: 12.4.254.20-electron.0
Have you tried closing and reopening the ide or resyncing the database?
I’ll try reopening it, but the issue is that I can’t be certain whether Cursor is accurately addressing my current code and providing responses based on it, or if it’s simply relying on the previous context to generate answers. This uncertainty makes me concerned that Cursor might not be working with the latest version of my code, which could lead to suggestions that disrupt the existing code structure. As a result, it’s difficult to fully trust its recommendations.
Is that a bug? can I do something to fix it?
You can start a new session, this should resolve the issue.
Thanks, but if I do that, will Cursor still have access to the same context?
No, the entire context needs to be added again, but you can ask the AI model in the existing session to summarize what you’ve done. It will provide a brief summary that you can add to the new context.
For reference, we are currently tracking down a bug where @ing a file in the Composer does not actually allow the AI to see it!
While we work on this, as recommended by Dean, the simplest workaround is to start a new Composer session if this happens, as that should enable the AI to see the file.