I am evaluating cursor as the value prop seems very compelling and I am interested in an IDE that has deeper AI integrations than most of the copilot extensions that I have tried.
Key features that are required for my team and I before this is considered more than a POC is file ignore list and auditable code to ensure that secrets are not being scraped from the codebase and being included in the embeddings.
If this is already in place I would love to see the implementation code, ideally this would integrate with the existing vscode exclusion patterns and .gitignore.
IMO this is a must have feature if it is not in place already.
Seems like the GH repo is just for issue management
I am under the impression that embeddings created from the indexed files will be persisted on a remote centrally managed db, is that the case or is there a way to only store embeddings locally or preferably provide an alternative gpt to create embeddings in the first place?
To verify the behavior you can add a file to the ignore list and try to talk about the file in Chat while using the @Codebase@-Symbol. The file shouldn’t appear in the references/context.
Yes, the embeddings are stored on Cursor’s managed database, and there’s currently no way to store embeddings locally or use a different model for them.
I can still reference them in the chat, even though I added them to .gitignore and .cursorignore, and resync index. I’ve confirmed this for sure in the devcontainer environment, and I believe it was the same on my Mac as well. What could be the issue?
