Precise Translation (Large Language Model)

An optimized high-performance large language model is built into the AI Pod, specifically designed for high-quality translation tasks, delivering accurate, fluent, and context-appropriate translations—ideal for scenarios with stringent translation quality requirements. Applications within the LCMD can directly invoke this service via the trans API exposed by the AI Pod.

Model Features

• High-Quality Translation: Powered by a large language model, delivering accurate and context-aware translations
• Multi-language Support: Supports mutual translation between major languages
• Flexible Configuration: Allows customization of translation requirements, such as language pairs and translation styles
• Strong Compatibility: Fully compatible with OpenAI API format, easing integration
• High-Performance Deployment: Optimized for deployment using the vLLM framework, with fast response times
• Excellent Foundation Model: Based on the Qwen3 model, delivering superior performance in translation tasks

Service Information

• API Endpoint:
• OpenAI-Compatible Endpoint:
• Default model name: aipod-trans (current version points to Qwen3 model; it is recommended to use alias aipod-trans for easier API calls and seamless future upgrades)
• Response mode: Supports both streaming and non-streaming responses
• Custom configuration: Supports specifying translation language, style, and other requirements via system prompt

Independent deployment usage

1. Save the following content to a docker-compose.yml file

services:
  trans:
    image: registry.lazycat.cloud/x/lzc-aipod-trans:3480c9d
    ports:
      - 3000:3000
    environment:
      - LANGUAGE=en_US:en
      - LANG=en_US.UTF-8
      - LC_ALL=en_US.UTF-8
      - VLLM_CONFIGURE_LOGGING=0
      - VLLM_DO_NOT_TRACK=1
      - VLLM_NO_USAGE_STATS=1
      - MAX_MODEL_LEN=2048
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:3000/health"]
      interval: 10s
      timeout: 30s
      retries: 100
      start_period: 1s

2. Run docker-compose up -d to start in the current directory
3. After startup, the http://127.0.0.1:3000/v1/chat/completions can be invoked