Global Options
Configure the AI provider, API keys, model, temperature and more.
Properties
The Kirby Copilot plugin can be configured globally in the config.php file of your Kirby project. All options are nested under the johannschopplich.copilot key.
return [
'johannschopplich.copilot' => [
// Customize Kirby Copilot
]
];
All possible options have reasonable defaults. Only if you want to change the default behavior, you need to add the option to your configuration.
The following list contains all available options:
provider String
The provider option defines the AI model provider that will be used to generate content across all Copilot sections. The plugin supports two providers:
openai(default)mistral
By default, the openai provider is used. To change the provider, you can use the following configuration:
return [
'johannschopplich.copilot' => [
'provider' => 'mistral'
]
];
providers Array
apiKey String
The providers option holds the configuration for each AI provider, such as the AI model to use for text generation. A valid API key is required for Kirby Copilot to make requests to the AI provider.
No API key is provided by default. Please be sure to add your API key to the configuration:
return [
'johannschopplich.copilot' => [
'providers' => [
'openai' => [
'apiKey' => 'YOUR_API_KEY'
]
]
]
];
model String
The model option defines the provider model used to generate content. In the case of OpenAI, the default configuration uses the gpt-4-turbo model:
return [
'johannschopplich.copilot' => [
'providers' => [
'openai' => [
'model' => 'gpt-4-turbo'
]
]
]
];
Migrating to Kirby Copilot v1.6+
As of April 2024, OpenAI has added vision capabilities to the GPT-4 Turbo model. Following this change, the configuration for the OpenAI provider has been simplified to only include the model name.To update your configuration from Kirby Copilot v1.5 and earlier to the latest version, reduce the model configuration to only include the model name:
return [
'johannschopplich.copilot' => [
'providers' => [
'openai' => [
- 'model' => [
- 'default' => 'gpt-4-turbo-preview',
- 'vision' => 'gpt-4-vision-preview'
- ]
+ 'model' => 'gpt-4-turbo'
]
]
]
];
For example, if you prefer the cheaper but less accurate gpt-3.5 model, you can change the default model:
return [
'johannschopplich.copilot' => [
'providers' => [
'openai' => [
'model' => 'gpt-3.5-turbo'
]
]
]
];
baseUrl String
The optional baseUrl option applies to the Mistral provider and lets you define the base URL of a custom Mistral API instance.
temperature Float
The temperature option defines the randomness of the generated content. The higher the temperature, the more random the generated content will be.
The default value is 0.5.
If you want the model to be more creative, you can increase the temperature:
return [
'johannschopplich.copilot' => [
'temperature' => 0.8
]
];
maxGenerationTokens Integer
The maxGenerationTokens option defines the maximum number of tokens that will be generated by the GPT model.
The default value is 1024.
If you want to generate an unlimited number of tokens, you can set the option to null.
return [
'johannschopplich.copilot' => [
'maxGenerationTokens' => null
]
];
systemPrompt String
The systemPrompt is probably the most important configuration option besides the model. It is the structural foundation of the generated content and complements the prompt that is provided by the user (userPrompt section property).
While the user prompt is intended to instruct the GPT model what to generate, the system prompt is intended to instruct the GPT model how to structure the generated content.
If you don't provide a systemPrompt, the plugin will use the following default prompt:
Provide responses in HTML format, including only the content that would go inside the <body> tags. Do not include the <head> section or any other parts of a full HTML document structure.
Note that the system prompt above works well for generating blocks or writer content, but it is not suitable for generating plain text content. If you want to generate plain text content, you should provide a system prompt that is suitable for plain text content:
return [
'johannschopplich.copilot' => [
'systemPrompt' => 'Provide responses in plain text without markup.'
]
];
blocksUpdateThrottle Integer
The blocksUpdateThrottle option defines the number of milliseconds that the plugin will wait before updating the blocks field with new generated content.
The default value is 250. Any value below 50 will be capped to 50.
A minimum throttle interval is required to avoid performance issues, because the HTML to blocks conversion is done via an internal API route.
Full Configuration Example
As a reference, below are all possible configuration options with their default values:
return [
'johannschopplich.copilot' => [
'provider' => 'openai',
'providers' => [
'openai' => [
'model' => 'gpt-4-turbo',
'apiKey' => 'YOUR_API_KEY'
]
],
'temperature' => 0.5,
'maxGenerationTokens' => 1024,
'systemPrompt' => 'Provide responses in HTML format, including only the content that would go inside the <body> tags. Do not include the <head> section or any other parts of a full HTML document structure.',
'blocksUpdateThrottle' => 250
]
];