Failed to load sessions - causing ram surge

[M01-Games]

Not sure what is going on as it has only just started happening but when I load my WebUI from my domain my ram usage goes to over 5GB when it would never hit 1GB before, also on my PC my session isnt loading but on my phone it is? ive got some images of errors and stats to show the issue, if you need anything else please let me know - would love any help as this project is amazing!

[AnnoDomine]

Hey! Don't worry, what you are seeing is actually completely normal behavior, but your PC errors point to a classic local browser stale state.

Why is the RAM Usage High?
What you are experiencing with the jump to over 5GB RAM is a performance optimization. The frontend of this WebUI is managed via Python static files. When started, it aggressively caches these files directly into system RAM, which is significantly faster than constant hard drive reads. As your served data grows, this number can jump suddenly, just as you described. It's safe to ignore as long as your system isn't running out of total RAM.

Why is the PC not Loading but the Phone is?
This difference is the key! If the phone connection works, it means your backend is serving data correctly and is accessible over the network. The problem is purely local to your PC's browser.

The specific errors you are seeing ("Failed to load sessions" and the 'Unexpected token <... is not valid JSON') are the smoking gun. This means when your PC's browser tried to request session data, it did not receive JSON data; it received an HTML error page (like a 404 or a stale authorization response) from your server instead. The browser parser got confused by the first character (<) and crashed. A stale auth cookie on the PC is the most common culprit.

Actionable Steps to Fix It:

• Try a Private/Incognito Window: This will immediately tell you if it's a caching issue.
• Clear Site Data/Cookies for this Domain: Clear all storage and cookies specifically for your domain on the PC. A stale session token is likely making the server reject the PC's request before the main application can even answer.
• Temporarily Disable Extensions: Adblockers or network privacy extensions on the PC can sometimes interfere with complex application routing in ways that mobile networks do not.

[M01-Games]

Just tried a private window and the first load worked but then after a refresh is still lost that session so not sure what is goin on there, then i cleared all cached data from my PC and Cloudflare domain and still on a normal browser it loaded first time and then after refresh is failed.
is there any way to just remove all saved data within files and start a fresh?
also the RAM spike never use to happen and it can cause crashes sometimes and thats with 8GB of DDR5 4800MHz SODIMM.
Any ideas on solutions for both

[AnnoDomine]

8GB of DDR5 4800MHz SODIMM

This right here is your problem. You are likely allowing Docker to use as much memory as possible out of those 8GB. But the OS needs memory too (on Windows mostly ~3-4GB).
You can investigate it by opening the Task Manager.
As the internal database grows over time (chat history, configs, user data, etc.), it requires more memory. Eventually, it caps out and completely fills your available RAM. So there is no space left, causing the crashes and failed loads.
Try to delete unused chats, for example, to reduce the needed memory footprint.

If you still want to completely start fresh like you asked and wipe all data, you can delete your Docker volumes – e.g., by running docker compose down -v – but be aware this will delete everything from the containers!

Another solution would be to increase the memory.

And if you run the model on the same machine, everything will break fast.

[M01-Games]

spare me another ai response please - like i said before this never happened before and it has 8GB of ram dedicated to its container that is running on my 32GB Linux NAS - also due to the caching issue i am doing to start a fresh install to ensure it wont happen again.

[AnnoDomine]

It doesn't matter if your NAS have 32GB or more or less.
If you gave docker only 8GB it only have 8GB and does not have any posibility to get more from the NAS.
Does it answer your question?

And if the NAS wants more than the the 24GB remains, it takes it from the container as it is the higher order system.
So if you model want to eat, it eats from the containers.