diff --git a/docs/my-website/docs/proxy/config_settings.md b/docs/my-website/docs/proxy/config_settings.md index 05e490a8b4..b518fe7da8 100644 --- a/docs/my-website/docs/proxy/config_settings.md +++ b/docs/my-website/docs/proxy/config_settings.md @@ -555,7 +555,7 @@ router_settings: | DEFAULT_MOCK_RESPONSE_COMPLETION_TOKEN_COUNT | Default token count for mock response completions. Default is 20 | DEFAULT_MOCK_RESPONSE_PROMPT_TOKEN_COUNT | Default token count for mock response prompts. Default is 10 | DEFAULT_MODEL_CREATED_AT_TIME | Default creation timestamp for models. Default is 1677610602 -| DEFAULT_NUM_WORKERS_LITELLM_PROXY | Default number of workers for LiteLLM proxy. Default is 4. **We strongly recommend setting NUM Workers to Number of vCPUs available** +| DEFAULT_NUM_WORKERS_LITELLM_PROXY | Default number of workers for LiteLLM proxy when `NUM_WORKERS` is not set. Default is 1. **We strongly recommend setting NUM_WORKERS to the number of vCPUs available** (e.g. `NUM_WORKERS=8` or `--num_workers 8`). | DEFAULT_PROMPT_INJECTION_SIMILARITY_THRESHOLD | Default threshold for prompt injection similarity. Default is 0.7 | DEFAULT_POLLING_INTERVAL | Default polling interval for schedulers in seconds. Default is 0.03 | DEFAULT_REASONING_EFFORT_DISABLE_THINKING_BUDGET | Default reasoning effort disable thinking budget. Default is 0 diff --git a/litellm/proxy/proxy_cli.py b/litellm/proxy/proxy_cli.py index 2bc1c8f8e9..2509a80b14 100644 --- a/litellm/proxy/proxy_cli.py +++ b/litellm/proxy/proxy_cli.py @@ -318,7 +318,7 @@ class ProxyInitializationHelpers: @click.option( "--num_workers", default=DEFAULT_NUM_WORKERS_LITELLM_PROXY, - help="Number of uvicorn / gunicorn workers to spin up. By default, it equals the number of logical CPUs in the system, or 4 workers if that cannot be determined.", + help="Number of uvicorn / gunicorn workers to spin up. Default is 1 (from DEFAULT_NUM_WORKERS_LITELLM_PROXY)", envvar="NUM_WORKERS", ) @click.option("--api_base", default=None, help="API base URL.")