# 作业17 · 生产 LiteLLM 主备路由配置（配置发布用）
# 与 patches_cursor/LiteLLM部署/config.yaml 对照；变更走 Git PR + 重启 litellm
#
# 发布：备份 config → 改本文件 → diff → 重启 → curl 验收 → 记审计
# 回滚：git checkout 上一版 config.yaml → docker compose restart litellm

model_list:
  # 虚拟模型：对外仍叫 qwen3.6-27b，内部可切主备 upstream
  - model_name: qwen3.6-27b
    litellm_params:
      model: openai/qwen3.6-27b
      api_base: http://116.62.10.183:8000/v1   # primary vLLM
      api_key: "123456"
      timeout: 300
      max_retries: 2
      extra_body:
        chat_template_kwargs:
          enable_thinking: false

  # 备模：同模型名不同 upstream（或降级小模型）
  - model_name: qwen3.6-27b-backup
    litellm_params:
      model: openai/qwen3.6-27b
      api_base: http://116.62.10.183:8001/v1   # backup 实例（示例端口）
      api_key: "123456"
      timeout: 300
      max_retries: 0

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

router_settings:
  routing_strategy: simple-shuffle

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY