Provider Configuration Reference

Detailed configuration reference for every provider. For step-by-step setup guides, see the Recipes overview.

Provider Index

Provider	File	Description
DeepSeek	deepseek.yaml	V4-Flash and V4-Pro via DeepSeek API
Azure Foundry	azure-foundry.yaml	GPT-5.4 family via Microsoft Foundry
NVIDIA	nvidia.yaml	DeepSeek, GLM, Kimi, Step via NVIDIA AI Endpoints
OpenCode Zen	opencode-zen.yaml	Free-tier models (Mimo, Nemotron, DeepSeek, MiniMax-M3)
OpenRouter	openrouter.yaml	Aggregated models via OpenRouter
OpenAI Compatible	openai-compatible.yaml	Any OpenAI-compatible /v1/chat/completions endpoint
Anthropic Compatible	anthropic-compatible.yaml	Any Anthropic-compatible /v1/messages endpoint
Local	local.yaml	Self-hosted inference servers on two ports

How a Provider YAML Works

Each provider YAML defines a model_list. Every entry has two parts:

model_list:
  - model_name: Provider/ModelName # Name used by aliases and clients
    litellm_params: # How to call the API
      model: deepseek/deepseek-v4-flash
      api_base: os.environ/DEEPSEEK_BASE_URL
      api_key: os.environ/DEEPSEEK_API_KEY
    model_info: # Capabilities and limits
      mode: chat
      context_window: 1000000

Source YAML files live at ~/.codefreedom/proxy/config/providers/ (after recipe install) or in the codefreedom-recipes repository.

Enabling a Provider

1. Uncomment the model entries in the provider YAML file.
2. Ensure the file is in the include list in config.yaml.
3. Set the API key in ~/.codefreedom/.env.proxy.secrets.
4. Restart the proxy: codefreedom run proxy restart.