Inline Completion Models

1 What inline completion does

Inline completion shows AI-generated ghost text while you edit Markdown or text files. It is useful for finishing a sentence, continuing a list, or drafting the next small part of a document.

When a suggestion appears:

• press Tab to accept it;
• press Escape to dismiss it;
• keep typing to replace it with your own text.

2 Open inline completion settings

Open the OpAgent app menu from the logo in the title bar, choose Models, then scroll to Inline completion.

3 Choose a mode

The completion mode controls which model is used.

• Auto — use OpAgent’s automatic inline-completion strategy when available, then fall back to the default model.
• Default — use the completion agent default model, then the global default model.
• Custom model — pin inline completion to one enabled model.
• Off — disable inline completion.

4 Use a dedicated completion model

1. Add or enable a model

   In Models, make sure the model you want is enabled. If it is not available yet, add it first from Add Custom Models.

2. Enable inline completion

   Turn on the Enable checkbox in the Inline completion section.

3. Choose Custom model

   Set the completion mode to Custom model.

4. Select the model

   Pick the model from the custom model dropdown.

5. Test in a document

   Open a Markdown or text file, place the cursor after some text, and wait briefly for a suggestion.

5 Which model should I use?

For inline completion, prefer a model that is fast and inexpensive. It does not need to be your strongest reasoning model.

Good completion models should:

• respond quickly;
• produce short continuations;
• follow the current document style;
• be cheap enough for frequent use.

Use a stronger model for chat and agent tasks when you need planning, reasoning, or complex edits.

6 If suggestions do not appear

Check these items first:

• Inline completion is enabled.
• Mode is not set to Off.
• The selected custom model is enabled.
• The file is a Markdown or text file.
• The editor has focus and the cursor is inside the document.
• The model provider is working.

If a custom model is slow, switch inline completion to a faster model and keep the stronger model for chat.