Semantic Kernel Tools: Prompt Engineering for Multiple Endpoints

Shannon Monroe

Are you curious how you can test your designs and applications against multiple endpoints, and compare cost or performance?  Are you ready to take your prompt engineering to the next level?

We are thrilled to announce an exciting new feature for Semantic Kernel Tools, the Visual Studio Code extension designed to enhance the development experience with Semantic Kernel. With the latest update, developers can now enjoy the benefits of managing multiple endpoints effortlessly. This feature empowers developers to compare AI system performance and seamlessly switch between different AI providers, making AI development more accessible and efficient than ever before.  Some of these new features include:

• Streamlined Endpoint Management: The Endpoint’s view takes center stage within the Semantic Kernel Tools extension, providing a user-friendly interface for developers to switch between different AI models quickly and easily. Whether you are working with Azure OpenAI, OpenAI, or Hugging Face, users can effortlessly navigate and select the desired endpoint.
• Azure OpenAI Integration:  By signing in with your Azure account, you gain access to the models deployed in your Azure subscription. The tree view presents Azure Subscriptions, Azure Cognitive Services (housing multiple model deployments), and Model Deployments. Enabling you to utilize text completion and chat completion models efficiently.
• OpenAI Integration: Semantic Kernel Tools seamlessly integrates with OpenAI, providing a hassle-free experience for exploring OpenAI models. Simply provide your OpenAI API key to unlock the vast array of models at your disposal. The tree view neatly organizes the models, categorizing them as either text completion or chat completion. With alphabetical sorting, finding the right model becomes a breeze.
• Hugging Face Integration: Hugging Face enthusiasts will be delighted by the seamless integration within Semantic Kernel Tools. No API key is required to access the list of available models. The tree view presents Hugging Face task types, including Text Generation and Text-to-Image. Additionally, a wide range of Hugging Face AI models are displayed. To ensure a smooth experience, the tree view conveniently loads the first 15 items, with the option to load more at the click of a button.

Enhanced Development Experience for AI Projects:

The new Multiple Endpoints feature enables developers to create, test, and compare AI models efficiently. By seamlessly switching between endpoints, developers can unleash the full potential of Semantic Kernel and tailor their projects to meet specific needs and goals.

Getting started is extremely easy:

1. In the AI Endpoints view, click on the arrow icon to switch the endpoints provider.
2. From the list of AI providers, select a new provider, such as HuggingFace.
3. You’ll have access to “Text Generation”   “Text to Image” and more.

To add Hugging Face Image Generation:

1. Create or open a semantic function and write your prompt.
2. In the AI Endpoints (Hugging Face) view, expand the “Text-to-Image” item and select an image generation model.
3. Run the function.  During the first execution, it will ask for an API key. Insert your Hugging Face API Key.

Semantic Kernel Tools has taken a significant step forward with the introduction of the Managing Multiple Endpoints feature. This game-changing update empowers developers to compare AI system performance, seamlessly switch between AI providers, and create innovative and impactful projects with ease.

Shannon Monroe

Follow