#Background Agents broken/blocked on one project

9 messages · Page 1 of 1 (latest)

whole haven
#

One of my projects is completely blocked for using Background Agents - either Cursor IDE or Web-based

I've recloned the repo twice, tried web, tried local, tried git hard reset, refetch, fsck - none of which reported any errors, but just to be doubly double sure...

*Allocating resources for this agent (request id: bc-4a8f8b77-c379-4d09-993a-0952eca7abc8)...

Error: Environment failed to start
Details:
fatal: You are attempting to fetch [commit-id], which is in the commit graph file but not in the object database.
This is probably due to repo corruption.
If you are attempting to repair this repo corruption by refetching the missing object, use 'git fetch --refetch' with the missing object.
fatal: could not fetch [different-commit-id-than-above] from promisor remote

[Status] Checking out repository…*

Other projects work fine, and local Cursor chats are just fine.

It's only the Cursor Background Agents that are breaking. Nothing special about project - just node.js/TypeScript code and standard package.json
The project is only about a week old, and has been working fine for a week until today - didn't do anything special other than start a new background agent

Seems like a Docker build cache, but I've used the 'clean' Docker/build option when launching the Background agent and it made no difference at all.

whole haven
#

ok what's the fix?

toxic brook
#

Hey there @whole haven my apologies if you’ve found that message coming along intrusive.

#

I’ll get back to you in a bit.

#

Interesting, based on what I could read here on the post this issue could be merged or send in #background-agent they will take care of this problem as I’ve read you have managed to fix some git bottlenecks but it turns out that it could be environment in general!

whole haven
#

I haven't been able to fix it - all of my research and attempts seem to imply its a problem in the background agent host environment - something presumably cached that it isn't flushing

#

it looks like a shallow clone of the main branch that got incorrectly pulled and isn't properly refetching

#

I have discovered that creating a branch off of main and targeting the background agent at that seems to work

whole haven
#

Actually scratch that - I definitely think it was broken in the remote agent env (Docker build/etc) but creating a branch manually - then asking the agent to operate in that one - then merging it back - has resolved main

I suspect because the separate branch forced the agent to do a deep or reclone (resetting its corrupted state) and then merging it back to main also caused a reset