Unable to chat with GGUF model. After downloading, duplicates and asks to download again.

[Algolgaming]

Describe the bug
AI Playground incorrectly constructs local file paths when attempting to add GGUF models from Hugging Face using the "Add Model" feature with the Llama.cpp-GGUF backend. This prevents users from loading GGUF models hosted on Hugging Face.

To Reproduce
Steps to reproduce the behavior:

Open AI Playground.
Select the Llama.cpp-GGUF backend.
Click the "+" button ("Add Model").
Enter a valid Hugging Face model identifier for a GGUF model, for example: Ttimofeyka/MistralRP-Noromaid-NSFW-Mistral-7B-GGUF/MistralRP-Noromaid-NSFW-7B-Q8_0.gguf
Click "Add."
Observe the error message dialog, which shows an incorrect local file path.
Expected behavior
AI Playground should recognize the input string as a Hugging Face model identifier, download the specified .gguf file from Hugging Face, and load the model into the Llama.cpp-GGUF backend. It should not attempt to find a local file.

Screenshots

The Models/ prefix being added to the path.
The use of backslashes \ instead of forward slashes /.
The resulting incorrect local file path.
Environment:
OS: [ Windows 11]
GPU: [Intel Arc A770 Bifrost 16G]
CPU: [ Intel i7-14700K]
Version: [AI Playground [v2.0.0-alpha-preview]
Additional context
The issue appears to be related to incorrect path handling within AI Playground. The input string, which is a valid Hugging Face model identifier, is being misinterpreted as a local file path. AI Playground is prepending resources/service\models\llm\ggufLLM\ (or similar) to the input and using backslashes \ instead of forward slashes /, resulting in an invalid local file path.

This issue is specific to adding GGUF models using the Llama.cpp-GGUF backend. It is not related to the transformers version issue that affects PyTorch models and the IPEX-LLM backend.

The incorrect path in the error message looks something like this:

resources/service\models\llm\ggufLLM\Ttimofeyka/MistralRP-Noromaid-NSFW-Mistral-7B-GGUF\MistralRP-Noromaid-NSFW-7B-Q8_0.gguf
The correct input provided by the user is:

Ttimofeyka/MistralRP-Noromaid-NSFW-Mistral-7B-GGUF/MistralRP-Noromaid-NSFW-7B-Q8_0.gguf
This clearly shows that AI Playground is constructing an incorrect local file path instead of using the provided string as a Hugging Face identifier.

This bug prevents the intended functionality of easily adding GGUF models from Hugging Face

[iMonZ]

Same issue...

Model: Qwen2.5-Coder-32B-Instruct-GGUF