LOCAL OR WEB LLM PROXY TRANSLATOR

[carlos-henrique-sa]

For English speakers this is not a problem, but for everyone else that writes prompts in other languages, it is. The idea would be a proxy, or something like that, that would use a LLM (being local or web) only to translate to English from any language, and translate back when the model returns data and texts. From what I have researched, this could cut out the token usage from 25% to 50%.

I already use something like that, but it is manual, copy paste, I use a local LLM to translate to English, using qwen 3.5 0.8b Q4KM on CPU. Would be nice if that could be incorporated into 9router. Thanks.