Feature Request: llama.cpp local endpoint support + multiple segmenters for multi-language

[IllIIllIlllIIl]

Hi, thank you for creating VectFox—it's a great idea and very useful for SillyTavern memory.

I have two feature requests related to my setup and language usage:

1. Support for llama.cpp local endpoint
   Right now VectFox supports VLLM and Ollama backends.
   My use case is

• Main roleplay model: runs on GPU, uses most of my VRAM
• Embedding/summarization model: I want this to run only on CPU/RAM

Because of that, llama.cpp is ideal for me: it allows small embedding models to run fully in RAM, without touching VRAM, so I can keep all VRAM reserved for the main RP model.

Feature request:

Add support for llama.cpp local endpoints as a backend option for VectFox.

This would be very helpful for users who need strict VRAM separation between RP models and vector/summarization models.

2. Multi-language support: allow using multiple segmenters
   On the settings page, the Multi-language support section currently lets me choose either Intl.Segmenter or Jieba WASM (TW dict)

My roleplay chats are a mix of Traditional Chinese and English, often in the same conversation.

For this kind of mixed-language usage, it would be ideal if VectFox could:

• Either use multiple segmenters at the same time (e.g. Intl.Segmenter + Jieba TW),
• Or provide a way to define per-language / per-field segmenter behavior.

Right now having to choose only one segmenter is a bit limiting for mixed Traditional Chinese + English scenarios.

If these are technically feasible, I think they would make VectFox much more flexible for multilingual roleplay setups and VRAM-constrained environments.

Thank you again for your work on this project!

[KritBlade]

1. It's on to do list, just too many people ask to have more end point. Eventually, vectfox will just use Sillytavern default connection profile. You define whatever profile in sillytavern, you can then just pick profile in vectfox on different embedding/summarizer/extractor

2. If you play in T-Chinese, you can just pick Jieba TW, it should work even if you have a mix of English and T-Chinese. English is always part of the dictionary even if you choose Jieba TW.