Skip to main content

Qwen Code

Qwen Code is Qwen's open-source command-line AI coding workflow tool. Official site: qwen.ai/qwencode

1.1 Usage​

Method 1: Config File​

Set your HPC-AI key in the shell:

export HPC_AI_API_KEY="your-hpc-ai-api-key"

Configure ~/.qwen/settings.json:

{
  "modelProviders": {
    "openai": [
      {
        "id": "minimax/minimax-m2.5",
        "name": "HPC-AI Minimax M2.5",
        "envKey": "HPC_AI_API_KEY",
        "baseUrl": "https://api.hpc-ai.com/inference/v1",
        "generationConfig": {
          "timeout": 120000,
          "maxRetries": 2,
          "samplingParams": {
            "temperature": 0.2,
            "max_tokens": 8192
          }
        }
      }
    ]
  },
  "security": {
    "auth": {
      "selectedType": "openai"
    }
  },
  "model": {
    "name": "minimax/minimax-m2.5"
  }
}

Then start Qwen Code:

qwen

Method 2: Runtime Flags​

Use runtime flags for a quick test before writing a persistent config:

export HPC_AI_API_KEY="your-hpc-ai-api-key"

qwen --auth-type openai \
  --model minimax/minimax-m2.5 \
  --openaiApiKey "$HPC_AI_API_KEY" \
  --openaiBaseUrl https://api.hpc-ai.com/inference/v1

1.2 References​