LocHub does not translate anything by itself. Every translation job sends strings to an AI provider you choose, using a translate model and a cheaper judge model that checks the translation afterwards. You pick the provider and both models in Project Settings > Plugins > LocHub, under the AI category (see Settings Reference for every field).
| Provider | Setting value | Default translate model | Default judge model |
|---|---|---|---|
| Anthropic (Claude) | Anthropic | claude-opus-5-5 |
claude-sonnet-5 |
| OpenAI | OpenAI | gpt-6-sol |
gpt-6-luna |
| xAI | xAI (Grok) | grok-4.7 |
grok-4.7 |
| DeepSeek | DeepSeek | deepseek-v4-pro |
deepseek-flash |
| Google Gemini | Google Gemini | gemini-3.8-flash |
gemini-3.5-flash-lite |
| Any OpenAI-compatible endpoint (Ollama, LM Studio, OpenRouter, Azure OpenAI, …) | Custom (OpenAI-compatible) | none — you set it | none — empty uses the translate model |
You can change either model to any model ID your provider's API accepts — the defaults above are just a sensible starting point (a cheaper model, or an older/newer generation).
Each provider has its own API Key field in Project Settings > Plugins > LocHub > AI, directly under AI Provider (and, for Anthropic specifically, under Anthropic Auth) — switching AI Provider shows that provider's own key field and hides the others. Paste the key in; it is masked while typing. For Custom (OpenAI-compatible) the key is optional — see the next section.
The API Key field under AI Provider
The key is saved in Config/DefaultEditor.ini along with the other LocHub settings, so it travels with the project like any other project setting, including through whatever version control the project uses: anyone who can read the project's config can read the key.
Changing the key applies to the running LocHub service the same way changing the provider or a model does: right away, or, while a translation job is running, right after that job finishes.
Custom (OpenAI-compatible) points LocHub at any server that speaks the OpenAI Chat Completions API: a model running on your own machine (Ollama, LM Studio, llama.cpp server, vLLM), a router such as OpenRouter, or a private deployment such as Azure OpenAI. With a local model your text never leaves your machine, and translating costs nothing per token.
Custom (OpenAI-compatible) selected in Project Settings
Set AI Provider to Custom (OpenAI-compatible), then fill in:
/chat/completions (for example http://localhost:11434/v1). It must start with http:// or https://; LocHub adds /chat/completions and /models itself. Logs and error messages only ever show its scheme, host and port.