OpenAI adapter

Uses the OpenAI-compatible Chat Completions API. MiniMax, Moonshot, DeepSeek, OpenRouter, and LM Studio use this adapter.

Provider section shape

Add the provider definition to your user provider catalog:

{
  "providers": {
    "my-provider": {
      "adapter": "openai",
      "enabled": true,
      "base_url": "https://llm.example.com/v1",
      "api_key_env": "MY_PROVIDER_API_KEY",
      "timeout_seconds": 60,
      "temperature": 1,
      "top_p": 1,
      "parallel_tool_calls": true,
      "tool_choice": "auto",
      "options": {},
      "models": [
        {
          "model": "my-model",
          "context_window_tokens": 128000,
          "max_output_tokens": 4096
        }
      ]
    }
  }
}

This example defines a new provider. When customizing a built-in provider, omit adapter and include only the fields you want to change.

Provider fields

Fields omitted from a built-in provider keep their catalog values; there is no single default shared by all providers.

Field	Description	Default
enabled	Makes the provider available.	From the included catalog
base_url	Chat Completions API base URL.	From the included catalog
timeout_seconds	Positive request timeout in seconds.	From the included catalog
temperature	Sampling temperature, zero or greater.	From the included catalog
top_p	Nucleus sampling probability, greater than 0 and at most 1.	From the included catalog
parallel_tool_calls	Requests support for parallel tool calls.	From the included catalog
tool_choice	Tool selection mode: auto, required, or none.	From the included catalog
options	Provider-specific body parameters, sent as extra_body.	From the catalog; empty object if not defined
models	Available models; replaces the complete inherited list.	From the included catalog

Credentials

Choose one API key source. Setting a source replaces the inherited source; do not set several in the same entry.

Field	Description	Default
api_key_file	Path to a private file containing only the API key.	Inherited key source
api_key_env	Environment variable containing the API key.	Inherited key source
api_key	Literal API key; prefer a file or environment variable.	Inherited key source

For example, configure MiniMax with a key file:

{
  "providers": {
    "minimax": {
      "api_key_file": "~/.skiller/secrets/minimax_api_key"
    }
  }
}

To configure credentials interactively, see Add an LLM provider.

Models

Each model entry declares its identifier and token limits. Defining models replaces the entire provider model list.

Field	Description	Default
model	Model identifier accepted by the provider.	Required for each model entry
context_window_tokens	Positive context window size in tokens.	Required for each model entry
max_output_tokens	Positive output token limit, sent as max_tokens.	Not set; no explicit limit sent

{
  "providers": {
    "minimax": {
      "models": [
        {
          "model": "MiniMax-M3.1-Flash",
          "context_window_tokens": 1000000,
          "max_output_tokens": 4096
        }
      ]
    }
  }
}

Provider options

Use options for body parameters specific to your service. For example, the included MiniMax definition enables reasoning_split:

{
  "providers": {
    "minimax": {
      "options": {
        "reasoning_split": true
      }
    }
  }
}

Only use parameters supported by the service. To define a new connection rather than customize a built-in provider, see Custom LLM providers.