Why is there llamacpp instead of ollama in its folder?

[Seedmanc]

The docs and app claim that Msty supports both, there's a clear division everywhere in the UI about Ollama and Llamacpp. But if you go to AppData\Local\Programs\MstyStudio\localai\lib\ollama there's the same llamacpp binaries instead. This is a misinformation, I was hoping to have an alternative to cpp.

[nikeshparajuli]

Hey @Seedmanc, Ollama tab shows items from Ollama's Library (https://ollama.com/library) whereas the GGUF are from HuggingFace. For that model specifically (https://huggingface.co/kaitchup/phi-1_5-safetensors), the model is in safetensors format, which we don't allow to pull directly from HuggingFace right now except for mlx models on mac. You can open a feature request on our discord server (http://msty.ai/discord). If you already have the model downloaded, you should be able to import it from the import tab.

The reason why you are seeing same llama cpp binaries is likely because Ollama switched to using llama cpp internally recently so they use the same files now. I'd recommend updating your engines to the latest version along with Msty to clear any confusion. cc @doponuraroco

We are enhancing Msty Studio significantly for 3.0. While the conversation experience will be improved in the versions after that, we aim to better organize things for model and settings pages. Please consider joining our discord server to provide additional feedback.

[nikeshparajuli]

msty_ollama_fix.zip Man, I ran into that same headache last week. Turns out the paths are just messy and it's pulling the wrong libs. I found a way to swap the symlinks so it actually calls the dedicated ollama engine instead of just looping back to the cpp files.

Also let us know which OS you are on and any relevant system and software (llama cpp, ollama, Msty Studio) version info.