litellm_config.yaml

model_list:
  - model_name: qwen
    litellm_params:
      model: nvidia_nim/qwen/qwen3.5-397b-a17b
      api_key: os.environ/NVIDIA_QWEN_API_KEY

  - model_name: glm5
    litellm_params:
      model: nvidia_nim/z-ai/glm5
      api_key: os.environ/NVIDIA_GLM5_API_KEY

  - model_name: minimax
    litellm_params:
      model: nvidia_nim/minimaxai/minimax-m2.1
      api_key: os.environ/NVIDIA_MINMAX_m2_1_API_KEY

  - model_name: step-3.5
    litellm_params:
      model: nvidia_nim/stepfun-ai/step-3.5-flash
      api_key: os.environ/NVIDIA_API_KEY

  - model_name: kimi-k2.5
    litellm_params:
      model: nvidia_nim/moonshotai/kimi-k2.5
      api_key: os.environ/NVIDIA_KIMI_2_5_API_KEY

  - model_name: glm4.7
    litellm_params:
      model: nvidia_nim/z-ai/glm4.7
      api_key: os.environ/NVIDIA_API_KEY

  - model_name: deepseek-v3.2
    litellm_params:
      model: nvidia_nim/deepseek-ai/deepseek-v3.2
      api_key: os.environ/NVIDIA_API_KEY

litellm_settings:
  # Drops unsupported parameters
  drop_params: true
  
  # --- PERFORMANCE OPTIMIZATIONS ---
  # Disables background telemetry to BerriAI
  disable_telemetry: true
  # Stops the proxy from writing every request/response to the server logs, 
  # which massively speeds up token streaming on free/weak CPUs.
  turn_off_message_logging: true

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY