Settings

Point the pipeline at any OpenAI-compatible API and pick your model. Keys are stored in this browser only (localStorage) and never sent anywhere except the endpoint you configure.

LLM engine

Provider preset

Choosing a preset fills the base URL and a default model. Pick "Custom" to enter your own.

Base URL

The root of an OpenAI-compatible API. /chat/completions and /models are appended automatically.

API key

Leave blank to keep the saved key.

Model

Click the field to browse every model the endpoint offers. Type to filter the list; Fetch models refreshes it.

Temperature

1.3 is the pipeline's tested sweet spot.

Output language

Translation hop — LibreTranslate (free)

Step 4 engine

No API key needed. Step 4 tries free LibreTranslate community mirrors (translate.disroot.org and friends) first, with the free Google endpoint as backup. Lingva, MyMemory and keyed engines (NiuTrans) were removed — dead or paid.

Intermediate language

The language your text passes through between Japanese and the output.