---
title: "Inline Suggestions"
description: "Ghost text appears as you type in writer fields – press Tab to accept, Escape to dismiss, or just keep typing to ignore."
canonical_url: "https://kirby.tools/docs/copilot/usage/inline-suggestions"
---
# Inline Suggestions
> Ghost text appears as you type in writer fields – press Tab to accept, Escape to dismiss, or just keep typing to ignore.
Pause typing in any writer field and Copilot suggests a continuation – press
to accept. Enabled by default for all writer fields – no configuration required.
Open the prompt dialog from the writer toolbar – generate or edit larger text blocks, with selection awareness.
Ghost text appears after a brief pause – press
to accept, keep typing to dismiss.
## How It Works
After a brief pause in typing (1 second by default), Copilot analyzes your current text and suggests a completion. The suggestion appears as ghost text in a subtle gray color after your cursor. Press
(Tab) to accept it,
to dismiss it – or just keep typing.
When a request to the provider fails, Copilot waits 30 seconds before suggesting again, rather than asking once more at every pause in typing. The manual trigger still works during that time – use it to retry immediately.
When you define **custom marks** in your writer field blueprint, you must explicitly include the `copilot-suggestions` mark to enable inline suggestions:
```yaml
fields:
text:
type: writer
marks:
- bold
- italic
# Append the Copilot toolbar button
- copilot
# Enable inline suggestions
- copilot-suggestions
```
## Keyboard Shortcuts
Action
Shortcut
Accept suggestion
(Tab)
Dismiss suggestion
Trigger manually
(macOS) /
(Windows/Linux)
By default, inline suggestions only trigger at the **end of a text node** – after a pause in typing. Use the manual trigger (
/
) to request a suggestion **between words** or in the middle of a sentence.
## Completion Model
Inline suggestions use a dedicated lightweight model optimized for speed. Configure the `completionModel` per provider in your [global configuration](/docs/copilot/configuration/global#default-models):
```php [config.php]
return [
'johannschopplich.copilot' => [
'provider' => 'google',
'providers' => [
'google' => [
'apiKey' => env('GOOGLE_API_KEY'),
'model' => 'gemini-3.1-pro-preview',
// Fast model for completions
'completionModel' => 'gemini-3.5-flash'
]
]
]
];
```
## Disabling Automatic Suggestions
To stop ghost text from appearing on its own in all writer fields:
```php [config.php]
return [
'johannschopplich.copilot' => [
'completion' => false
]
];
```
The manual trigger keeps working – press
whenever you do want a suggestion. To take inline suggestions out of a field altogether, leave the `copilot-suggestions` mark out of its blueprint.
## Customizing Debounce Timing
The debounce controls how long Copilot waits after you stop typing before suggesting a completion. Customize the debounce duration (in milliseconds) in your configuration:
```php [config.php]
return [
'johannschopplich.copilot' => [
'completion' => [
'debounce' => 1500 // Wait 1.5 seconds
]
]
];
```
The default debounce is 1000ms (1 second). Values below 500ms are clamped to 500ms to prevent excessive API calls.
---
Every page of this site as Markdown: