Tilburg.ai

Tilburg University's AI assistant for VS Code: a chat panel and inline code completions, both backed by tilburg.ai's OpenAI-compatible API.

Features

• Chat — ask questions from a dedicated panel in the Activity Bar, with Markdown rendering, code blocks with a Copy button, and a model picker.
• Edit Mode — toggle Edit Mode under the chat input and Tilburg.ai suggests changes directly in the attached file instead of answering in chat. The suggestion opens as a diff; Accept applies it (undoable with Ctrl+Z), Reject discards it.
• Inline code completions — ghost-text suggestions as you type, in any language. Toggle, disable, or snooze them from the Tilburg.ai status bar item in the bottom right (click it for the menu), just like GitHub Copilot's status icon.
• Model selection — switch between the models configured in config.yaml via the "Tilburg.ai: Select Model" command or the toolbar dropdown.

Getting started

1. Install the extension.
2. Run Tilburg.ai: Set API Key (Command Palette, or the button in the Tilburg.ai panel) and enter your tilburg.ai API key. It's stored in VS Code's Secret Storage — never written to a file or synced anywhere else.
3. Open the Tilburg.ai panel from the Activity Bar to start chatting, or just start typing in any file to see inline completions.

Configuration

Model settings live in config.yaml. Drop a config.yaml in your workspace root to override the bundled default; each model entry's roles array controls what it's used for:

models:
  - name: Tilburg gpt-5.4-nano
    provider: openai
    model: gpt-5.4-nano
    apiBase: https://api.tilburg.ai/v1
    roles: [chat, edit, apply, autocomplete]

Privacy

Chat messages and, for inline completions, the code surrounding your cursor are sent to the tilburg.ai API (configured via apiBase above) to generate responses. No code or conversation content is sent anywhere else, and nothing is retained locally beyond the current VS Code session.

Commands

Requirements

VS Code 1.86 or later. A tilburg.ai account and API key.