Autocomplete

Kilo Code's autocomplete feature provides intelligent code suggestions and completions while you're typing, helping you write code faster and more efficiently. It offers both automatic and manual triggering options.

How Autocomplete Works

The extension uses Fill-in-the-Middle (FIM) completion powered by Codestral (mistralai/codestral-2508). It analyzes the code before and after your cursor to generate contextually accurate inline suggestions.

Triggering Options

Auto-trigger

Autocomplete is enabled by default and automatically shows inline suggestions as you type. Suggestions appear as ghost text that you can accept with Tab.

Trigger on keybinding (Cmd+L)

Press Cmd+L (Mac) or Ctrl+L (Windows/Linux) to manually request a completion at your cursor position.

đź“ťNote

This keybinding requires kilo-code.new.autocomplete.enableSmartInlineTaskKeybinding to be enabled in VS Code settings. It is disabled by default.

Status Bar

The extension displays an autocomplete status indicator in the VS Code status bar, including:

  • Current autocomplete state (active/snoozed)
  • Cumulative cost tracking for autocomplete requests

Snooze / Unsnooze

You can temporarily disable autocomplete by clicking the status bar item to snooze it. Click again to unsnooze and re-enable suggestions.

Copilot Conflict Detection

The extension automatically detects if GitHub Copilot inline suggestions are enabled and warns you about potential conflicts. Disable Copilot's inline completions for the best experience with Kilo Code autocomplete.

Best Practices

  1. Balance speed and quality: Faster models provide quicker suggestions but may be less accurate
  2. Adjust trigger delay: Find the sweet spot between responsiveness and avoiding too many API calls
  3. Use Quick Task for complex changes: It's designed for more substantial code modifications
  4. Use Manual Autocomplete for precision: When you need suggestions at specific moments
  5. Configure providers wisely: Consider using faster, cheaper models for autocomplete while keeping more powerful models for chat

Tips

đź’ˇTip

When to use chat vs autocomplete: Use chat for multi-file changes, refactoring, or when you need to explain intent. Use autocomplete for quick, localized edits where the context is already clear from surrounding code.

đź’ˇTip

Steer autocomplete with comments: Write a comment describing what you want before triggering autocomplete, or type a function signature—autocomplete will fill in the implementation.

đź’ˇTip

Treat suggestions as drafts: Accept autocomplete suggestions quickly, then refine. It's often faster to fix a 90% correct suggestion than to craft the perfect prompt.

  • Autocomplete works best with clear, well-structured code
  • Comments above functions help autocomplete understand intent
  • Variable and function names matter - descriptive names lead to better suggestions