williampepple1/ibani-translator-backend

# Ibani Translator Backend This is a Node.js Express backend for the Ibani Translator, using the Hugging Face model williampepple1/ibani-translator. ## Prerequisites - Node.js (v16 or higher) - A Hugging Face API Token (Recommended for Inference API) ## Getting Started 1. **Install Dependencies:** ```bash npm install ``` 2. **Setup Environment Variables:** Create a `.env` file in the root directory and add your Hugging Face Access Token: ```env PORT=5000 HUGGING_FACE_ACCESS_TOKEN=your_token_here ``` 3. **Run the Server:** - Development mode: `npm run dev` - Production mode: `npm run start` ## API Endpoints ### 1. Health Check - **URL:** `/health` - **Method:** `GET` - **Success Response:** `{"status": "UP", "model": "williampepple1/ibani-translator", ...}` ### 2. Translate - **URL:** `/api/translate` - **Method:** `POST` - **Body:** ```json { "text": "I eat fish" } ``` - **Success Response:** ```json { "original_text": "I eat fish", "translated_text": "A mine í-njí ríerí", "source_language": "English", "target_language": "Ibani", "success": true } ``` ## Local Execution with Transformers.js (Optional) If you wish to run the model locally on your machine without an API key, you can use `@xenova/transformers`. Note: This requires the model to have ONNX weights. If they are not available, the Inference API approach is the most reliable. ## Model Details - **Base Model**: `Helsinki-NLP/opus-mt-en-mul` - **Language Pair**: English → Ibani - **Developer**: William Pepple