Adding AI in your apps with the Windows Copilot Runtime APIs

The latest versions of the Windows App SDK (1.7 Experimental 3 and 1.8 Experimental 1) introduce a powerful new feature to elevate your apps: the Windows Copilot Runtime (WCR) APIs. This set of APIs lets you integrate AI into your apps without the hassle of downloading models, fine-tuning them, or working with complicated interfaces.

With the WCR APIs, you can perform AI-powered language queries using just three lines of code:

using var languageModel = await LanguageModel.CreateAsync();
var response = await languageModel.GenerateResponseAsync("What's the formula for glucose");
Console.WriteLine(response.Text);

Pretty simple, right? And that’s just the beginning: you can use various built-in models to add AI capabilities to your apps:

• Phi-Silica – a language model for GPT-style queries
• OCR – optical character recognition that detects text in images, including lines, words, and bounding boxes
• Image Enhancer – improves image resolution while preserving detail
• Background Remover – removes the background from images, letting you select which parts to keep
• Image Description – generates descriptive text for images
• Object Eraser – removes objects from images and fills in the background seamlessly

One thing to note: these models are only available on Copilot+ PCs equipped with a Neural Processing Unit (NPU). If you have one of these devices and want to dive deeper, check out the the WCR website, or explore the AI Dev Gallery I mentioned in my last post.

The AI Gallery offers hands-on samples and example code to help you start using these APIs in your own projects. Happy AI coding!