# 207 LiteLLM 切流样例：主上游 → K8s Ingress；备上游 → 现网 116 Docker vLLM
# 对照：patches_cursor/LiteLLM部署/config.yaml、作业17 主备路由
# 变更仍走：备份 → diff → 拷到 /opt/litellm → systemctl restart litellm → curl 验收

model_list:
  - model_name: qwen3.6-27b
    litellm_params:
      model: openai/qwen3.6-27b
      api_base: http://vllm-qwen.gpu-k8s.example.local/v1
      api_key: "sk-cluster-gpu"
      timeout: 300
      max_retries: 2
      extra_body:
        chat_template_kwargs:
          enable_thinking: false

  # 回滚锚点：现网 116 Docker vLLM
  - model_name: qwen3.6-27b-docker116
    litellm_params:
      model: openai/qwen3.6-27b
      api_base: http://116.62.10.183:8000/v1
      api_key: "123456"
      timeout: 300
      max_retries: 0

  - model_name: qwen-small
    litellm_params:
      model: openai/qwen-small
      api_base: http://vllm-small.gpu-k8s.example.local/v1
      api_key: "sk-cluster-gpu"
      timeout: 120
      max_retries: 1

litellm_settings:
  drop_params: true
  set_verbose: false
  request_timeout: 300
  num_retries: 2
  fallbacks:
    - qwen3.6-27b: [qwen3.6-27b-docker116]

router_settings:
  routing_strategy: simple-shuffle

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY