Completions

Ghost-text completions

As you type, xpreiIDE suggests a completion inline (greyed-out “ghost” text) — press Tab to accept it, like any other inline suggestion. Toggle this on or off with the completions.enabled setting.

If your configured completion model is a FIM-trained code model (codellama, deepseek-coder, qwen2.5-coder, codestral, codegemma, granite-code, starcoder, or similar), xpreiIDE automatically uses Ollama’s native fill-in-the-middle completion for genuinely better suggestions — no setup needed beyond picking that model. Any other model falls back to a chat-based completion approach that works with anything, just at somewhat lower quality.

Commit message generation

In the Source Control panel, with some changes staged, click Generate Commit Message. The model reads your staged diff and writes a commit message into the input box for you to review, edit, and commit as usual.