Open mhnikravan-trilogy opened 1 month ago
The problem is bigger and not only in th codebase that never works as expected.
Even when you point a @folder and ask it to searching for specific code, if you check for context it is using all shorts of files out of this folder.
This is a serius security issue, especially for a project that decided to be closed source.
Have absolutely the same problem here. Before i was able to get the context of the codebase, i was connecting through ssh to my remote server from the windows machine.
I have exactly the same issue. I am also on WSL2. It appears to index the codebase (takes about a minute to get to 100%, it is showing all the filenames while doing it). But when it gets to 100%, a chat @codebase does not give any context (not using @codebase, not using the button). Removing the sync, resyncing - none of it helps.
I figured I'd go through the output to find any error messages. Hopefully they can be of assistance in fixing the issue.
Main:
2024-06-12 10:47:13.403 [info] update#setState disabled
2024-06-12 10:47:13.404 [info] update#ctor - updates are disabled as there is no update URL
2024-06-12 10:47:13.417 [error] [storage state.vscdb] error checking size of src.vs.platform.reactivestorage.browser.reactiveStorageServiceImpl.persistentStorage.workspaceUser: Error: Unexpected number of rows - 0
2024-06-12 10:47:13.418 [error] [storage state.vscdb] error checking size of workbench.panel.aichat.view.aichat.chatdata: Error: Unexpected number of rows - 0
Search++:
2024-06-12 10:47:18.061 [info] Backend URL: https://api2.cursor.sh
2024-06-12 10:47:18.202 [info] created the client
2024-06-12 10:47:18.202 [info] Is file sync enabled: false
Cursor indexing and retrieval:
2024-06-12 10:47:18.061 [info] Repo client: syncing embeddings with backend: https://repo42.cursor.sh
2024-06-12 10:47:18.254 [info] Creating merkle client.
2024-06-12 10:47:18.254 [info] Done creating merkle client.
2024-06-12 10:47:18.254 [info] Doing a startup handshake.
2024-06-12 10:47:18.329 [info] Finished initializing merkle tree in 75.51348500000313 ms.
2024-06-12 10:47:18.330 [info] Doing the initial handshake with hash: a02072e70d79ad963dc22918155e825032c132971b4ff2364aee28c80bfbaef2
2024-06-12 10:47:18.347 [info] Handshake start
2024-06-12 10:47:18.776 [info] Handshake timing: 445.28441199986264
2024-06-12 10:47:18.776 [info] Handshake result:
2024-06-12 10:47:18.776 [info] Starting fast remote sync.
2024-06-12 10:47:18.777 [info] Total num embeddable files: 464
2024-06-12 10:47:18.779 [info] Root hash: a02072e70d79ad963dc22918155e825032c132971b4ff2364aee28c80bfbaef2
2024-06-12 10:47:18.779 [info] Finished indexing repository.
2024-06-12 10:47:18.779 [info] Indexing job successfully done or aborted.
2024-06-12 10:57:18.255 [info] Creating merkle client.
2024-06-12 10:57:18.255 [info] Done creating merkle client.
2024-06-12 10:57:18.255 [info] Doing a startup handshake.
2024-06-12 10:57:18.320 [info] Finished initializing merkle tree in 65.53679799987003 ms.
2024-06-12 10:57:18.321 [info] Doing the initial handshake with hash: a02072e70d79ad963dc22918155e825032c132971b4ff2364aee28c80bfbaef2
2024-06-12 10:57:18.328 [info] Handshake start
2024-06-12 10:57:18.742 [info] Handshake timing: 420.773186000064
2024-06-12 10:57:18.742 [info] Handshake result:
2024-06-12 10:57:18.742 [info] Starting fast remote sync.
2024-06-12 10:57:18.743 [info] Total num embeddable files: 464
2024-06-12 10:57:18.743 [info] Root hash: a02072e70d79ad963dc22918155e825032c132971b4ff2364aee28c80bfbaef2
2024-06-12 10:57:18.743 [info] Finished indexing repository.
2024-06-12 10:57:18.743 [info] Indexing job successfully done or aborted.
Upon asking chat a question about the codebase, I simply get:
Gathering chunks using conversation history (... keeps spinning, shows no output)
Reranking chunks (... keeps spinning, shows no output)
To provide an accurate response, I need to see the code in question. Could you please share the relevant code snippet?
I am going to cancel my subscription as I do not get the value without being able to use the codebase context.
Same problem here. Works fine if I copy the repo over to my Windows file system and open it there, but @codebase context is completely broken with a WSL:Ubuntu workspace. It seems to be able to evaluate the file I have focus on, and that's it unless I go through all the other files and do ctrl-shift-L to add them to the chat. Also, the apply button for suggested changes doesn't work like 90% of the time. It just opens a new file tab. Seems like I used to see it work occasionally before, but seems totally broken now. This sucks, and I will totally cancel my pro subscription and warn others about it rather vocally. Cursor folks - please respond.
The codebase context is not working for me. I am a Cursor Pro subscriber and this feature seems to be the biggest advantage of Cursor AI to me (vs copilot, etc). So I appreciate your feedback on it.
1) If you can, please include a screenshot of your problem:
and the indexing is on and has been done on my code:
![image](https://github.com/getcursor/cursor/assets/166420546/02cba5d1-94e9-46a8-ad43-4674b10a183f)
Even if I change the![image](https://github.com/getcursor/cursor/assets/166420546/8f6d01e4-f53f-4943-9801-6e42eee713a5)
include
to be something like*.*
or*.md
still nothing, while I know the information is there:2) Please include the name of your operating system I am using "Microsoft Windows 11 Home", and running cursor on WSL.
![image](https://github.com/getcursor/cursor/assets/166420546/00a4af30-db49-4f17-b357-35aaf804c78d)
3) If you can, steps to reproduce are super helpful.
On another note: By using the
codebase
context, is it supposed to use all the indexed codebases (i.e., whatever codebases I have ever indexed) or just the codebase in the current workspace? If it's the first one, how can I change the behavior to focus on the codebase in my current workspace? I work with lots of repositories, and using everything that has been indexed can easily misguide the AI.