Implementation:BerriAI Litellm Proxy Config Loader

Knowledge Sources	Domains	Last Updated
BerriAI/litellm repository	LLM Gateway, Configuration Management, Proxy Infrastructure	2026-02-15

Overview

Concrete tool for loading and applying YAML-based proxy configuration provided by the LiteLLM ProxyConfig class.

Description

The ProxyConfig class is the central abstraction for all configuration loading and updating logic in the LiteLLM proxy server. It reads a YAML configuration file, resolves environment variables, applies litellm module-level settings (such as caching, callbacks, and parameter behavior), constructs or reconfigures the litellm.Router with declared model deployments, and applies general server settings. The load_config method returns a tuple of the initialized router, the model list, and the general settings dictionary, which are then assigned to the proxy server's global state.

The class also supports:

Validating whether a file is a YAML file via is_yaml().
Loading raw YAML content from disk via _load_yaml_file().
Retrieving the merged config from file or database via get_config().
Maintaining internal state of the loaded config for subsequent reads.

Usage

Import and use ProxyConfig when initializing or reconfiguring the LiteLLM proxy server. It is instantiated as a module-level singleton (proxy_config) and called during the proxy_startup_event lifespan handler or when a configuration reload is triggered.

Code Reference

Attribute	Value
Source Location	`litellm/proxy/proxy_server.py`, class defined at line 2021
Method	`ProxyConfig.load_config()`, defined at line 2459
Signature	`async def load_config(self, router: Optional[litellm.Router], config_file_path: str) -> Tuple[Router, list, dict]`
Import	`from litellm.proxy.proxy_server import ProxyConfig`

I/O Contract

Inputs

Parameter	Type	Description
`router`	`Optional[litellm.Router]`	An existing Router instance to reconfigure, or `None` if a new Router should be created.
`config_file_path`	`str`	Filesystem path to the YAML configuration file to load.

Outputs

Return Element	Type	Description
`router`	`litellm.Router`	The initialized or updated Router instance containing all model deployments and routing settings.
`model_list`	`list`	A list of model deployment dictionaries, each containing `model_name`, `litellm_params`, and `model_info`.
`general_settings`	`dict`	A dictionary of server-level settings extracted from the configuration (e.g., master key, database URL, auth modes).

Usage Examples

Basic configuration loading during server startup:

from litellm.proxy.proxy_server import ProxyConfig

proxy_config = ProxyConfig()

# During the server startup lifespan event
llm_router, llm_model_list, general_settings = await proxy_config.load_config(
    router=None,
    config_file_path="/app/config.yaml"
)

Example YAML configuration file (config.yaml):

environment_variables:
  OPENAI_API_KEY: "os.environ/OPENAI_API_KEY"

model_list:
  - model_name: gpt-4
    litellm_params:
      model: openai/gpt-4
      api_key: os.environ/OPENAI_API_KEY
  - model_name: claude-3
    litellm_params:
      model: anthropic/claude-3-opus-20240229
      api_key: os.environ/ANTHROPIC_API_KEY

litellm_settings:
  drop_params: true
  cache: true
  cache_params:
    type: redis
    host: localhost
    port: 6379

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY
  database_url: os.environ/DATABASE_URL

router_settings:
  routing_strategy: least-busy
  num_retries: 3

Reloading configuration at runtime:

# Reconfigure existing router with updated config
llm_router, llm_model_list, general_settings = await proxy_config.load_config(
    router=existing_router,
    config_file_path="/app/config.yaml"
)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment