Feature Request/Bug Report

[disturbedlegend]

Here's my bug / feature list:

• Incorporate folder hierarchy to the left-side recent chat history so that users can organize them (for example, 'work', 'work/project x', 'work/project y', 'personal')

• In addition to the Assistant 'edit' and 'delete' functions, add 'duplicate' (so minor variations don't require re-typing all of the custom elements or hacking the JSON files)

• When selecting a model from the 'New Chat', 'Hi, how are you?' page, longer model names are cropped in the drop-down list, making it impossible to discern model variants (e.g. Qwen3-Coder-30B-A3B-Instruct-1M-Q4_K_M.gguf from Qwen3-Coder-30B-A3B-Instruct-1M-UD-Q8_K_XL.gguf) as hovering does not work while selecting, only after a selection is made. This is particularly frustrating when Font Size is set to Large or Extra Large.

• Incorporate dynamic determination of the number of model layers to offload to the GPU - if the entire model fits, great, but if it doesn't extend what doesn't fit to the CPU. Current support for a static assignment is useless because there's no simple way to assess the GB/layer cost to extrapolate how many layers to offload.

[louis-jan]

@disturbedlegend We implemented Project feature so that you can create multiple Projects now.
There is also Auto-Optimized model settings to determine NGL layers could be used.

[disturbedlegend]

Jan team,

I frequently operate with dozens for Assistants that tune model parameters for different use cases, which makes for a convoluted mess when trying to find a particular assistant - can you add the ability to sort or re-order assistants?

Thank you,