UnityGameTranslator Files

What's New

Universal AI API (OpenAI-compatible)

• Any LLM provider: Ollama, LM Studio, OpenAI, Groq, Google Gemini, Mistral, OpenRouter, and any OpenAI-compatible server
• API Key support: Encrypted at rest (AES-256) for cloud providers
• Model dropdown: Browse available models from your server with searchable dropdown + Refresh button
• Backward-compatible config: Existing Ollama configs auto-migrate to new format

Thinking Model Support

• Universal think: false for all models that support it
• Extra /no_think + assistant prefill for Qwen3/DeepSeek-R1 models
• Non-thinking models (TranslateGemma, Llama, etc.) get clean requests

Bug Fixes

• Fix Font.GetOSInstalledFontNames() crash on IL2CPP (Schedule I, etc.)
• Fix silent AI errors — HTTP failures and exceptions now always logged
• Fix config migration not triggering worker thread on first load
• Reduce log spam: verbose font diagnostics moved behind debug_ai flag
• Silence expected ForceMeshUpdate failures on uninitialized components

Configuration

:::json
{
  "ai_url": "http://localhost:11434",
  "ai_model": "qwen3:8b",
  "ai_api_key": null,
  "enable_ai": true
}

For cloud providers (e.g., Groq):

:::json
{
  "ai_url": "https://api.groq.com/openai",
  "ai_model": "llama-3.3-70b-versatile",
  "ai_api_key": "your-api-key-here",
  "enable_ai": true
}