Getting started

Install BdebTech AI IDE, download the Starter model and get your first offline completion in about three minutes.

What you get

BdebTech AI IDE is a VS Code-style editor (built on VSCodium) with three AI features that run entirely on your computer:

  • Tab autocomplete - grey ghost text as you type; press Tab to accept.
  • Chat - a sidebar assistant that can see the current file or your selection.
  • Edit with AI - describe a change to selected code, review a diff, apply or discard.

The intelligence comes from an open-weight model (Qwen2.5 Coder) served by llama.cpp through a small local daemon called bdebd. There is no account, no API key, and no network traffic after the one-time download.

1. Install

Download the Windows installer from the download page and run it. It installs into your user profile (no administrator rights needed) and creates a Start menu entry.

The preview build is not code-signed yet. If Windows SmartScreen appears, choose More info → Run anyway.

2. First launch

  1. Open BdebTech AI IDE from the Start menu.
  2. A notification asks to download the Starter model (Qwen2.5 Coder 0.5B, 645 MB). Click Download now.
  3. The status bar shows progress. The daemon also downloads the matching llama.cpp engine for your hardware (about 20 MB on CPU).
  4. When it finishes, the status bar reads Qwen 0.5B and the engine starts.

Everything lives in %LOCALAPPDATA%\BdebTech\AI (models, engines, logs, config). Uninstalling the IDE keeps this folder unless you choose to delete it.

3. First completion

Open any code file (or create demo.ts) and type:

export function isEven(n: number): boolean {
  return

Stop typing for a moment. Ghost text appears: n % 2 === 0;. Press Tab.

On a 2021 Intel i5 laptop with no GPU this takes roughly half a second. On a modern desktop it is near-instant.

4. Chat and edits

  • Ctrl+Alt+L opens the chat sidebar. Tick Include current file to attach the file; a selection is attached automatically.
  • Right-click selected code → BdebTech AI → Explain, Fix / Improve, Generate Tests.
  • Ctrl+Alt+K with code selected → type an instruction → review the diff → Apply.

5. Go bigger (optional)

Click the model name in the status bar → Switch model. The list marks what your hardware can run. Lite+ (1.5B) is free and noticeably smarter on 8 GB+ machines. Studio models (7B-32B) need a Studio license and ideally a GPU.

Next