Sugoi Toolkit and Sugoi 14B/32B

Edit User-Settings.json to point to your local instance (LM Studio: localhost:1234, oobabooga: localhost:5000 or Ollama: localhost:11434).

LM studio

lm_studio/<model name>, server http://127.0.0.1:1234/v1 and key sk-xxx.

Note 1:  lm_studio/ is absolutely required in the model_name string when loading a model, lm_studio/ tells litellm which backend driver to use

Note 2:  /v1 absolutely required in api_server eg http://127.0.0.1:1234/v1

Note 3:  key also absolutely required to be written in api_key and not be left empty, with a real or fake key eg sk-xxx .

Oobabooga

oobabooga/model name, server http://127.0.0.1:5000 without any key.

Ollama

ollama/model name, server http://127.0.0.1:11434 without any key.

Note: Sugoi uses internal system prompt which optimized for cloud llm (Translation-API-Server/LLM/Translator.py). Might have accuracy hit on offline model.

Run Sugoi Toolkit

• Open sugoi toolkit and click on Sugoi Translator LLM

• Wait until the server is ready

• Copy your text

Note: Streaming output currently unsupported by the Sugoi Japanese Translator frontend.