Want to build a Copilot for your app? Semantic Kernel & Prompt Flow for Beginners

[Guest MatthewAnderson]

Hi there, I’m back with another post, I recently co-presented on a developer-focused webinar on Semantic Kernel for beginners. This webinar was part of a series about leveraging Azure OpenAI, with this one diving into how to add intelligent experiences into your new and existing applications.

 

 

In the webinar, we focused on the AI Orchestration layer of the Copilot Stack, which is a framework that helps you design and implement AI solutions using Azure OpenAI. We focused on two tools: Semantic Kernel and Azure Prompt Flow. Semantic Kernel is an SDK that allows you to interact with AI models using natural language commands, while Azure Prompt Flow is a tool that helps you create and manage prompts for AI models.

 

 

The focus of my portion of the session (starts at 25:28 in the video) was geared toward developers and the use of the Semantic Kernel SDK to easily add AI into applications. I showed how you can use the SDK in both C# and Python to create plugins, plans, and personas for Semantic Kernel.

 

• Plugins - modules that connect Semantic Kernel to different AI models or services, such as GPT-3, Azure Prompt Flow, or the Microsoft Graph.
  
• Plans - sequences of actions that Semantic Kernel can execute to achieve a goal, such as booking a flight or writing some content.
  
• Personas - profiles that define the preferences and personality of Semantic Kernel, such as tone, style, and humor.
  

 

I used a demo to introduce these concepts, where I deconstructed the Bing Chat extension for the Edge browser, going through development of a simplified version. I used Semantic Kernel to handle the natural language understanding and generation, as well as to access Deepa’s web service (which she created in Prompt Flow, see below) through plugins.

 

 

 

 

My co-presenter Deepa covered the use of Azure Prompt Flow, which is a graphical interface that lets you create and edit prompts for AI models. Deepa showed how you can use Azure Prompt Flow to create prompts from scratch or from templates, test them on different models and datasets, and deploy them to Semantic Kernel or other applications.

 

I thoroughly enjoyed presenting this session and I hope you learned something new and useful from it. If you missed it or want to watch it again, you can find the recording on demand on YouTube. You can also check out the other webinars in the series about Azure OpenAI.

 

 

 

Want to learn more? Here are the "Resources" links that we included toward the end of the webinar:

 

• Discord Community – active conversations with weekly office hours Join the Semantic Kernel Discord Server!
  
• Semantic Kernel on GitHub – source code, backlog, sample apps, notebooks
  
• Microsoft Learn docs – quick start, tutorials, hackathon materials Orchestrate your AI with Semantic Kernel
  
• Chat Copilot sample – Build your own integrated large language model chat copilot GitHub - microsoft/chat-copilot
  
• Semantic Kernel recipes – Start cooking today https://aka.ms/sk/recipes
  
• Microsoft Learn doc - https://aka.ms/AML-PF
  
• RAG Pattern with prompt flow - RAG using PF
  

 

Continue reading...