how to update the lora listing cache in swarm after deleting safetensors in file system? swarm doesnt seem to upated correctly to reflect the true files in path after deletion. Is there a best practise for Swarm to see only the remaining files? Refresh button does not rescan the paths
#How to update Lora (cache) listing after deleting from file system
dapper sentinel
lost ledge
I have noticed this being an issue for the last several weeks as well.
dapper sentinel
I have noticed this being an issue for the last several weeks as well.
Currently looking for information in past help posts.
Are you saying this is a bug in recent swarm?
I think it is more likely it is a tricky roundabout way to clear and regenerate the cache
dapper sentinel
Reset metadata?
dapper sentinel
IN UTILITIES:
Metadata Utilities
Reset All Metadata
If you click this button, all Model and Image metadata datastores will be reset, and they will be reloaded from source files.
This is useful for example if you've externally modified your model or image files and want to clean out or update Swarm's metadata tracking.
Note that this does not remove metadata from the models or images themselves, just from the databases that keep an efficient tracker of them.
This may take a moment to run.
from research:
Metadata Retention Behavior
Preserved Elements
Embedded Model Metadata
- CivitAI-style trigger phrases in SAFETENSORS - headers26
- Preview images stored as PNG payloads in model files
- Architectural descriptors (e.g., SD1.5 vs SDXL classification)
User-Created Content
- Custom thumbnails saved via Attach Preview feature
- Descriptions/notes added through Swarm's metadata editor
- Trigger phrase modifications using native tools
Reset-Affected Elements
- Derived Cache Data
- In-UI search indexes
- Model relationship graphs
Batch generation presets using metadata filters6
- Transient States
- Recent usage counters
- Dynamic popularity rankings
- Session-specific ordering preferences
Ok works. Closed and opened after reset
golden blaze
how to update the lora listing cache in swarm after deleting safetensors in file...
uhhh what
you should not see a file in the model list after hitting Refresh if the model was deleted
it might be stuck in cache somewhere but cache is not what populates the model list, just metadata about models in the list
I'm not aware of any bugs in this regard
lost ledge
Started seeing some stuck models or loras within the last two weeks or so. After fully deleting and confirming they were not on the file system anymore (in 2 GPU back ends) they still show in the list after a refressh and meta reset.
dapper sentinel
I'm not aware of any bugs in this regard
Could be Dropbox locking or OneCommander
lost ledge
Are you using multiple back ends by chance Rafraf?
dapper sentinel
Are you using multiple back ends by chance Rafraf?
Sometimes. But not seeing this issue at the moment. Only noticed on local machine and remote machine
lost ledge
I may actually started to see this issue after I started using multiple GPU back ends.
golden blaze
Could be Dropbox locking or OneCommander
... you have your models in dropbox? wowza. Yeah that'll have weird side effects
golden blaze
Started seeing some stuck models or loras within the last two weeks or so. Afte...
is it two comfy self starts, or do you have a swarm-api-backend? with swarm-api-backend, it grabs a model list from the other instance and combines local+remote lists
lost ledge
Swarm-api-backend on a remote machine.
Even when I confirmed the file was removed from the file system for both backends it still stayed in the list.
golden blaze
it might not be properly forwarding the refresh request to the remote backend?
ie only updating the local list and not the remote one
im pretty sure it should work right tho
lost ledge
how do I debug, first let me make sure I can replicate the issue.
1: Primary UI I delete model, refresh list (Still shows). 2: Model is deleted from primary, not secondary. 3: Manually delete the model from secondary, refresh primary web UI, still see the model int he list.
golden blaze
what if you hit refresh again
lost ledge
Still shows after a refresh
Is it pulling from the second backup cache? Do I need to login there and refresh on that end?
golden blaze
blink
a typo has been sitting there and not reporting errors due to an unrelated mistake for who knows how long
golden blaze
fix pushed, should be all good now
lost ledge
confirmed fixed 🚀
dapper sentinel
i can confirm it is working now for me too. though i may have been inadvertantly deleting from file system instead of fro UI oops
dapper sentinel
@golden blaze for all DROPBOX users, keeping their cloud storage lora's synced to their local machine for mirroring changes to metadata and safetensors, the best practise I have found is deleting lora's on dropbox website, rather through UI or local file system. This bypasses lock files set by dbx. Also, important to notice the lora preview in UI tab will not refresh if you still have Lora on remote backend. Once remote is updated, it all falls back flush on both sides. Nice to know Swarm works nicely if file management is dealt with on dropbox web files side