Inline AI
Ghost Suggestions
Ghost Suggestions are AI-generated inline code completions that appear as you type — similar to GitHub Copilot. They are powered by your currently selected model and appear automatically after a brief pause.
How it works
After you stop typing for a moment (250ms by default), UltraCoder sends the current file context to the AI model and displays the suggestion as greyed-out ghost text directly in the editor. Press Tab to accept it.
Keyboard controls
| Action | macOS | Windows / Linux |
|---|---|---|
| Accept suggestion | Tab | Tab |
| Accept all suggestions | ⇧Tab | Shift+Tab |
| Dismiss suggestion | Esc | Esc |
| Next suggestion | ↓ | ↓ |
| Previous suggestion | ↑ | ↑ |
| Prompt for suggestion | ⌘I | Ctrl+I |
✦ Tip:Press ⌘I (or Ctrl+Ion Linux) to explicitly prompt for a suggestion at the cursor position, even if you haven't just typed anything.
Enabling and disabling
Ghost Suggestions are on by default. To toggle them:
- Open the Command Palette (⌘⇧P) → Preferences: Open Settings (UI)
- Search for
ultra-coder.inlineCompletion.enabled - Toggle the checkbox
Settings
| Setting | Default | Description |
|---|---|---|
ultra-coder.inlineCompletion.enabled | true | Enable or disable Ghost Suggestions globally |
ultra-coder.inlineCompletion.debounceMs | 250 | Milliseconds of typing silence before a suggestion is requested (50–2000) |
ultra-coder.inlineCompletion.maxTokens | 150 | Maximum tokens for each suggestion (50–1000). Higher = longer completions but slower. |
ℹ Note:Ghost Suggestions use the same model selected in the AI panel. Fast models (like
usf-mini) give lower-latency suggestions. If you're using a slow or rate-limited model, increase debounceMs to reduce unnecessary requests.Related
- Code Actions — right-click Explain, Fix, Improve on selected code
- Models & Providers — change the model powering suggestions