Skip to content
llms.py logo llms.py

Configuration

Configuration Files

Section titled “Configuration Files”

llms.py uses two main configuration files stored in ~/.llms/:

  • llms.json - Provider and model configuration
  • ui.json - Web UI settings and system prompts

These files are automatically created with defaults when you run llms --init.

llms.json Structure

Section titled “llms.json Structure”

The main configuration file has the following structure:

{
  "defaults": {
    "headers": {},
    "text": {},
    "image": {},
    "audio": {},
    "file": {},
    "check": {},
    "limits": {},
    "convert": {}
  },
  "providers": {}
}

Defaults Section

Section titled “Defaults Section”

Headers

Section titled “Headers”

Common HTTP headers for all requests:

"headers": {
  "Content-Type": "application/json"
}

Text Template

Section titled “Text Template”

Default chat completion request for text prompts:

"text": {
  "model": "kimi-k2",
  "messages": [
    {"role": "user", "content": ""}
  ]
}

Image Template

Section titled “Image Template”

Default request template for image prompts:

"image": {
  "model": "gemini-2.5-flash",
  "messages": [
    {
      "role": "user",
      "content": [
        {"type": "image_url", "image_url": {"url": ""}},
        {"type": "text", "text": "Describe the key features of the input image"}
      ]
    }
  ]
}

Audio Template

Section titled “Audio Template”

Default request template for audio prompts:

"audio": {
  "model": "gpt-4o-audio-preview",
  "messages": [
    {
      "role": "user",
      "content": [
        {"type": "input_audio", "input_audio": {"data": "", "format": "mp3"}},
        {"type": "text", "text": "Transcribe the audio"}
      ]
    }
  ]
}

File Template

Section titled “File Template”

Default request template for file attachments:

"file": {
  "model": "gpt-5",
  "messages": [
    {
      "role": "user",
      "content": [
        {"type": "file", "file": {"filename": "", "file_data": ""}},
        {"type": "text", "text": "Summarize the document"}
      ]
    }
  ]
}

Providers Section

Section titled “Providers Section”

Each provider has the following structure:

"groq": {
  "enabled": true,
  "type": "OpenAiProvider",
  "base_url": "https://api.groq.com/openai",
  "api_key": "$GROQ_API_KEY",
  "models": {
    "kimi-k2": "moonshotai/kimi-k2-instruct-0905",
    "llama3.3:70b": "llama-3.3-70b-versatile"
  },
  "pricing": {
    "kimi-k2": {"input": 0.0, "output": 0.0}
  },
  "default_pricing": {"input": 0.0, "output": 0.0}
}

Provider Fields

Section titled “Provider Fields”
  • enabled: Whether the provider is active
  • type: Provider class (OpenAiProvider, GoogleProvider, OllamaProvider)
  • api_key: API key (use $VAR_NAME for environment variables)
  • base_url: API endpoint URL
  • models: Model name mappings (local name → provider name)
  • pricing: Cost per 1M tokens (input/output) for each model
  • default_pricing: Default pricing if not specified in pricing

Managing Providers

Section titled “Managing Providers”

Enable/Disable Providers

Section titled “Enable/Disable Providers”
Terminal window
# Enable providers
llms --enable openrouter groq google_free


# Disable providers
llms --disable ollama openai

List Providers

Section titled “List Providers”
Terminal window
# List all providers and models
llms ls


# List specific providers
llms ls groq anthropic

Set Default Model

Section titled “Set Default Model”
Terminal window
llms --default grok-4-fast

This updates defaults.text.model in your configuration.

Custom Configuration Path

Section titled “Custom Configuration Path”

Use a custom configuration file:

Terminal window
llms --config /path/to/custom-config.json "Hello"

Recreating Configuration

Section titled “Recreating Configuration”

To reset to defaults, delete your configuration:

Terminal window
rm -rf ~/.llms
llms --init

Next Steps

Section titled “Next Steps”