Auto-Function Calling with Semantic Kernel

Publicado em: 17 Dezembro 2024
no canal de: Brett Garnier

158

Github Repo with grounding files for Windsurf:

https://github.com/bgarnr/WindsurfDocs

Microsoft Semantic Kernel Repo:

https://github.com/microsoft/semantic...

Microsoft Learn

https://learn.microsoft.com/en-us/dot...

*Description:*

In this video, I introduce and demonstrate **Semantic Kernel's auto-function calling**, focusing on how developers familiar with Microsoft technologies can integrate AI into their applications.

Key Concepts Introduced and Covered:

*Auto-Function Calling:*
I show how this feature allows an AI application to dynamically decide which function to call during a chat conversation, mapping language to actions. This enables a more conversational and flexible user experience, allowing the system to update the UI or backend dynamically based on user input.

*Semantic Kernel:*
I discuss this Microsoft library, which facilitates auto-function calling and much more. It’s the core technology featured in this video.

*Use Case:*
I demonstrate a narrative fiction writing application that uses Semantic Kernel. The demo highlights auto-function calling by setting a world JSON object and saving it to Entity Framework Core using natural language input.

---

Practical Tips and Recommendations:

*Starting Point:* I recommend visiting the Semantic Kernel's GitHub repo and Microsoft Learn for helpful resources.
*File Sharing:* I’ve provided a GitHub repo link below with the files I used in this demonstration alongside WinSurf.
*WinSurf Tool:* I highly recommend using WinSurf, especially in combination with Semantic Kernel.
*File Organization:* To streamline architecture and design discussions, copy files in different entity folders into a single file and upload them to Claude.
*Version Matching:* Make sure your NuGet version of Semantic Kernel matches the version used in my GitHub repo to avoid compatibility issues.
*Injecting Chat ID:* Use the chat ID in Semantic Kernel to better control your chat messages.
*Prompt Engineering:* Plan ahead, as the values pulled from plugins will be part of your prompt.
*Plugins:* I distinguish between reading and writing to a database using plugins.
*Log Everything:* Log all function calls to Semantic Kernel for better tracking and troubleshooting.
*Choice Behavior Auto:* When setting up Semantic Kernel, set the Function Choice Behavior to *auto* so the AI can dynamically choose which function to call.
*Streaming:* Use streaming to display large text responses or summaries while sending function calls in a single message.

---

Workflow Discussion:
I describe a workflow where I code primarily in *WinSurf* for quick actions, such as generating or checking code snippets with AI.

For longer, high-level discussions about product goals or platform architecture, I suggest using *Claude* to have more strategic and design-focused conversations.

---

Thank you for watching! Don’t forget to check out the GitHub repo link above and leave any questions or thoughts in the comments.

Nesta página do site você pode assistir ao vídeo on-line Auto-Function Calling with Semantic Kernel duração hora minuto segundo em boa qualidade , que foi baixado pelo usuário Brett Garnier 17 Dezembro 2024, compartilhe o link com seus amigos e conhecidos, no youtube este vídeo já foi visto 158 vezes e gostou 7 espectadores. Boa visualização!

2,725