Codebase indexing causing cursor to freeze up repeatedly?

Bug Reports
1

Anyone else having an issue that the codebase index tries to resync and EVERY time it does it makes cursor become unresponsive until it finishes? it started occurring on this last update for me, in literally any and every single workspace and project that i have. Even my smallest ones -.- almost unusable unless i disable indexing

2

update, 110% its an issue with the new update. not sure if its even codebase index causing the issue but i downgraded back to 0.47 and i dont have any issues now. 0.49 does not seem stable what so ever.

note to self; never agree to using the early access updates and gotta love having every single chat history totally wiped

3

Hi, Cursor dev here. Can you give a little more information about what’s going on, and how to reproduce this?

Specifically:

  • What platform are you on?
  • Are your workspaces stored locally, on a networked drive, or on a different computer?
  • What specifically makes you think indexing is the problem? It freezes every time you click “Resync” on the settings page?
  • Are you able to reproduce on 0.48?
  • Is there anything interesting in the “Cursor Indexing & Retrieval” logs (in the bottom pane, where the terminal is, you can go to the Output tab and choose those logs from the dropdown menu to see them).
  • Would you be willing to do “Capture and send debugging data” from the cmd/ctrl-p menu so we can check out specific performance data for you?

Thanks, and sorry for the bad experience.

4

Windows
Locally.
Seems like whenever its syncing up the filebase or something it triggers and freezes.
Oddly enough its changed to just crashing frequently in general. In all my work spaces, I’ve tried switching back to .47, but the terminal crashes all the time and i have to reload dev window constantly.
0.48 has an array of issues.
0.49 crashes probably every 15 minutes or every time I open it a few times.

Worst part is, I lose my chats for projects betwen every time I switch versions so this whole week or two has been a mess lol.

I ran capture and send debugging data a few times. Dunno exactly how that works.

Thanks for the response, its not like you all intentionally provide bad experiences. Shame I didn’t response sooner, may have helped myself and others if I was more prompt about it.

Anything else I can do to help?

Edit: to clarify, 0.47 had the terminal crashing issue prior to upgrading but I’m used to it. Problem is, every time I downgrade and leave cursor running it updates itself to 0.49 and wipes my chats so I can’t even stick with the functional version. I’ve had to resort to running everything externally in my CMD instead of in cursor or I quite literally cannot get anything done at all. Its also crashing when using long running commands like backtesting for trading then it ends up losing the output instead of showing the agent the error at the end; hence externally having to run it.

A lot of the time, the command terminal runs but it doesnt fully expand to show the output from the running command. So I have to either stop yolo mode and re run it or I have to go back and run it again manually.

1 Like
5

I have the same issue, it is driving me nuts. It keeps throwing Windows for a loop and raising the “has stopped responding” system warning, and I tell it to wait, and it starts working again, but it happens roughly every 10 minutes.

6

I have the same issue. We got ripped off, it doesnw work, nor support and no refund.

7

I wouldn’t go that far as saying it’s a rip off. I’ve been using Cursor for awhile now and been happy with it. Just right now it is particularly unstable

8

Yeah. I’m not saying I’m super unhappy with cursor in general either. Always workarounds to the functionality, but I cannot say it’s not irritating sometimes lol. When I first got cursor, it was super smooth, no issues, models were great - seemingly no weird temperature like issues.

Now it’s like the random updates that barely add or change functionality do anything significant enough to call for the frequent changes. Had a good product, but being force updated is off putting in itself. It’s wild to think that minor - SUPER minor updates call for this many pushes alongside the fact that it barely adds any significance what so ever a majority of the time.

To be honest, I don’t think there’s a single update over the past 5 versions that I’ve cared for what so ever. All I’d ask for is model temperature control (Claude is a spastic reeree space cadet compared to anthropics base claude idk what cursor does to it, and you can tell Gemini is far from deterministic so the temperature is too high too) and for chat to be consistently backed up and to transfer across versions instead of needing spec story :joy:

Heck, if a developer sees this; PLEASE for the love of god or whatever deity you follow is - add a Project Manager Instance, and Worker Instance modes along with actually beneficial MCPs by default and your customer base will grow significantly.

The chain of thought process is also dog poopie so I’ve heavily modified mine and it’s significantly better than the base models lol.

You’re making a product for A.) good developers or B.) vibe coders. Your product should support both yet we’re entirely lacking in project management or any useful degree of context persistence - both of which are things that are insanely easy to fix and to provide at the base of your product..

9

This is still happening (bursts of unresponsiveness out of the blue and Windows panicking about it), please please fix. It is getting more frequent now, I am not sure what it is. I’m now having to drop and restart my Cursor every half hour because it starts to flip out doing whatever it is doing in the background. A happy hack here would just letting me disable whatever is happening. I am not confident it is codebase indexing either. I am also not sure if this is a WSL2 thing either because my codebase that is on the windows partition is also showing this same issue.

The one hint I see is that there’s a suspicious amount of disk activity and pops of gpu activity. So, perhaps it is indexing, but why! I’ve not added any new files! I’m clearly not idling. Chrome is using an enormous amount more of memory than Cursor and I still have surplus, so it isn’t that either. One thing that is particularly odd about its footprint is that it is maintaining 81 processes under its belt, and all of them seem to be active on the CPU despite me focusing only on two of them. VSCode didn’t have this behavior, so something was added that was clever and a little too eager.

10

back up all your rules, custom modes, track your settings, extensions, themes, icon packs, etc. anything and everything that is custom in cursor aside from the base installation.

go into your app data → roaming → cursor. delete the entire folder

this should make a significant difference, if not entirely fix the issue.

Happy hunting

11

That worked, somewhat. It is still lagging a lot when I am just scrolling on a single file (<800 lines! Not even that long!). It is still freezing up spontaneously, but not so bad that windows gets worried. I just hope the next version they are hyping has a patch for whatever issue they introduced in this version.

Thanks, but I think the problem ultimately remains. It must be doing some sort of indexing over history and getting lost in it.

12

Ctrl+Shift+P and type ‘send’ then select ‘Developer: Capture and Send Debugging Data’

This will help the cursor devs identify and work on solving this problem in future iterations. Not enough users do this but everyone that has these issues can contribute with this manner. It might be best to send the data several times when the issues are occurring, which sounds frequent enough in your case.

Apologies I could not solve the issue fully, wish I could help further.

Happy hunting

1 Like
13

Nevermind, the improvement lasted only one day or so. It is crashing again, exact same behavior as before I cleared the files. I am not sure, maybe my codebase is too big, or I am doing something extremely unusual unknowingly. When I try to send performance data it also fails. (“Failed to start performance profile”)

I hope it all resolves with the new version.

14

Try enabling beta tests on updates, maybe 0.50 will work better. That or downgrade to an earlier version through the website.

How big of a codebase are we talking?

15

One is 50,000 or so, another is pushing closer to 180k LLoC. Not unreasonable sizes at all tbh, just “large” compared to what I am seeing vibe coded lately.

16

Do you use SpecStory extension?

17

No, I don’t know what that is. I’m pretty light on prompting outside of the cursorules and copious documentation files.

18

Are you loading up an enormous amount of those documentation files?

Mainly python library specific?

It may be better and less of a load if you leave an explicit instruction to use the help function and have the agent write a script that provides a text file output that is specific to the errors or issues its resolving or leveraging from those libraries. Then it can parse the output file.

Here is an example:

Script for Help in VectorBT Library, specific to IndicatorBase attributes :)

import vectorbt as vbt

with open(‘vbt_help_output.txt’, ‘w’, encoding=‘utf-8’) as f:
f.write(‘— help(vbt) —\n’)
# help(vbt) # This can be too verbose, let’s try specific parts

f.write('\n--- dir(vbt) ---\n')
f.write(str(dir(vbt)) + '\n')

if hasattr(vbt, 'indicators'):
    f.write('\n--- dir(vbt.indicators) ---\n')
    f.write(str(dir(vbt.indicators)) + '\n')
    if hasattr(vbt.indicators, 'base'):
        f.write('\n--- help(vbt.indicators.base) ---\n')
        # help(vbt.indicators.base) # Might be too long for one file read
        f.write('\n--- dir(vbt.indicators.base) ---\n')
        f.write(str(dir(vbt.indicators.base)) + '\n')
        if hasattr(vbt.indicators.base, 'IndicatorBase'):
             f.write('\n--- Found vbt.indicators.base.IndicatorBase ---\n')

if hasattr(vbt, 'base'):
    f.write('\n--- dir(vbt.base) ---\n')
    f.write(str(dir(vbt.base)) + '\n')
    if hasattr(vbt.base, 'indicator'): # This was the problematic one
        f.write('\n--- dir(vbt.base.indicator) ---\n')
        f.write(str(dir(vbt.base.indicator)) + '\n')
        if hasattr(vbt.base.indicator, 'IndicatorBase'):
             f.write('\n--- Found vbt.base.indicator.IndicatorBase ---\n')
    if hasattr(vbt.base, 'IndicatorBase'): # Direct check
        f.write('\n--- Found vbt.base.IndicatorBase ---\n')

# Attempt to import common modules and check for IndicatorBase
try:
    from vectorbt.indicators.base import IndicatorBase
    f.write('\n--- Successfully imported from vectorbt.indicators.base import IndicatorBase ---\n')
except ImportError:
    f.write('\n--- Failed to import from vectorbt.indicators.base import IndicatorBase ---\n')

try:
    from vectorbt.base import IndicatorBase
    f.write('\n--- Successfully imported from vectorbt.base import IndicatorBase ---\n')
except ImportError:
    f.write('\n--- Failed to import from vectorbt.base import IndicatorBase ---\n')

try:
    from vectorbt.indicators import IndicatorBase
    f.write('\n--- Successfully imported from vectorbt.indicators import IndicatorBase ---\n')
except ImportError:
    f.write('\n--- Failed to import from vectorbt.indicators import IndicatorBase ---\n')

print(‘Help script created.’)

Text File Output's Results

— help(vbt) —

— dir(vbt) —
[‘ATR’, ‘AlpacaData’, ‘ArrayWrapper’, ‘AsyncJob’, ‘AsyncScheduler’, ‘AtomicConfig’, ‘BBANDS’, ‘BOLB’, ‘BinanceData’, ‘CCXTData’, ‘CacheCondition’, ‘CancelledError’, ‘Config’, ‘Configured’, ‘Data’, ‘DataUpdater’, ‘Drawdowns’, ‘EntryTrades’, ‘ExitTrades’, ‘ExpandingSplitter’, ‘FIXLB’, ‘FMAX’, ‘FMEAN’, ‘FMIN’, ‘FSTD’, ‘Figure’, ‘FigureWidget’, ‘GBMData’, ‘IndicatorFactory’, ‘LEXLB’, ‘Logs’, ‘MA’, ‘MACD’, ‘MEANLB’, ‘MSTD’, ‘MappedArray’, ‘NumbaExperimentalFeatureWarning’, ‘OBV’, ‘OHLCSTCX’, ‘OHLCSTX’, ‘Orders’, ‘Portfolio’, ‘Positions’, ‘RAND’, ‘RANDNX’, ‘RANDX’, ‘RPROB’, ‘RPROBCX’, ‘RPROBNX’, ‘RPROBX’, ‘RSI’, ‘RangeSplitter’, ‘Ranges’, ‘Records’, ‘Rep’, ‘RepEval’, ‘RepFunc’, ‘RollingSplitter’, ‘STCX’, ‘STOCH’, ‘STX’, ‘ScheduleManager’, ‘SignalFactory’, ‘Sub’, ‘SyntheticData’, ‘TRENDLB’, ‘Trades’, ‘YFData’, ‘builtins’, ‘cached’, ‘doc’, ‘file’, ‘loader’, ‘name’, ‘package’, ‘path’, ‘pdoc’, ‘spec’, ‘version’, ‘_settings’, ‘_typing’, ‘_version’, ‘atomic_dict’, ‘base’, ‘cached_method’, ‘cached_property’, ‘data’, ‘deep_substitute’, ‘generic’, ‘import_submodules’, ‘indicators’, ‘labels’, ‘make_figure’, ‘make_subplots’, ‘merge_dicts’, ‘messaging’, ‘nb’, ‘ohlcv_accessors’, ‘pandas_ta’, ‘plotting’, ‘portfolio’, ‘px_accessors’, ‘records’, ‘returns’, ‘root_accessors’, ‘save_animation’, ‘set_seed’, ‘settings’, ‘signals’, ‘symbol_dict’, ‘ta’, ‘talib’, ‘utils’, ‘warnings’]

— dir(vbt.indicators) —
[‘ATR’, ‘BBANDS’, ‘IndicatorBase’, ‘IndicatorFactory’, ‘MA’, ‘MACD’, ‘MSTD’, ‘OBV’, ‘RSI’, ‘STOCH’, ‘all’, ‘builtins’, ‘cached’, ‘doc’, ‘file’, ‘loader’, ‘name’, ‘package’, ‘path’, ‘pdoc’, ‘spec’, ‘whitelist’, ‘basic’, ‘configs’, ‘factory’, ‘nb’, ‘pandas_ta’, ‘ta’, ‘talib’, ‘tp’]

— dir(vbt.base) —
[‘ArrayWrapper’, ‘all’, ‘builtins’, ‘cached’, ‘doc’, ‘file’, ‘loader’, ‘name’, ‘package’, ‘path’, ‘pdoc’, ‘spec’, ‘accessors’, ‘array_wrapper’, ‘column_grouper’, ‘combine_fns’, ‘index_fns’, ‘indexing’, ‘reshape_fns’]

— Failed to import from vectorbt.indicators.base import IndicatorBase —

— Failed to import from vectorbt.base import IndicatorBase —

— Successfully imported from vectorbt.indicators import IndicatorBase —

19

Naw, its C and Javascript, no Python in either of them. The editor freezes up even when I am not using chat/composer, just coding with occasional autocompletes. But let me try this idea of yours, I like it.

Also to answer the other question, no I just have the tabs open of files I am working on, and when I call in composer I’ll @ tag the files I need it to be aware of.