I am in the process of evaluating Cursor AI and it’s been working great for a few days. This morning, I decided to see how well it could create a README.md for an old project.
I open the folder, created a new workspace then I simply prompt the Composer with
“Review all the files in the root folder and use your knowledge of .NET to create a READM.md file that describes the solution.”
This has worked fine for other solutions, but this morning I get this response from the Composer.
" I don’t see any files or code from a solution in our conversation yet. To help analyze your project, you’ll need to share the relevant files or code with me. You can either:
- Share the code directly in our conversation
- Share file names and their contents
- Share a link to a public repository (though I may not be able to access external links)
Once you share the code, I’ll be happy to:
- Index all the files
- Analyze the code structure
- Provide insights about the project architecture
- Help with any specific questions or changes you need
Please share the code you’d like me to look at!"
I simply need the ability to navigate to a new solution and Cursor should have the ability to build a new index for the workspace and this should provide contextual awareness specifically for that solution.
Why is this not working? I DO NOT want to edit copy or have to @TAG files that is too much work. ← if this is the case I will go back to “Aider or Cline”.
Any help on how to resolve this is appreciated!
Hey, this can happen in long Composer sessions. Starting a new session should solve this problem.
I just worked in VS Community edition for a couple hours then when I reopened Cursor the Composer displayed a message saying "Note: codebase not indexed yet. Results are much better when indexed.
My question is this. I work on a lot of projects and clearly the embeddings and index are very important. How do I manage the index and can I have separate indexes per repo?
Just trying to better understand how Cursor manages this. I mean I do not want one index for all projects as I am sure Cursor is using a semantic similarity lookup then injecting those documents into the context for evaluation from the LLM
Each folder you open is separately indexed, and you are right, Cursor is smarter with a fully complete index.
You can see the status of the index in your Cursor settings (once you have a folder open), but it’s usually a pretty boring page, indexing your codebase the first time you open it then periodically reindexing as it goes.
If you open the parent folder of some code, containing multiple projects, it is reindexed separately so you never have any bleed between projects if you open them individually in Cursor.
Regarding the AI not seeing your files, this depends on how you use it. In normal composer mode, you have to use the @ symbol or the + button to add context for the Ai to see, whereas in agent mode, the Composer can lookup the context itself, so you don’t have to give it context manually.