Recap for April 2nd, 2025 - Semantic Kernel Office Hours for US/EMEA

[sophialagerkranspandey]

Semantic Kernel Office Hours for US/EMEA - April 2nd, 2025 recording: https://aka.ms/SKOfficeHoursRecordings

If you weren't able to attend office hours but still have questions, please post them here: https://github.com/microsoft/semantic-kernel/discussions

Meeting notes:
• RRC Release: Shawn announced the release of RRC, which includes updates and bug fixes. He encouraged attendees to review the updated documentation and provide feedback.
○ RRC Announcement: Shawn announced the release of RRC, which includes updates and bug fixes. He emphasized the importance of reviewing the updated documentation and encouraged attendees to provide feedback.
○ Documentation Updated: Shawn mentioned that all documentation and samples have been updated. He encouraged attendees to review the docs and report any issues they encounter.
○ Feedback Requested: Shawn expressed interest in receiving feedback on changes and improvements, and emphasized the importance of having a stable framework that is ready to use in applications.
• Semantic Agent Framework: Shawn discussed the Semantic Agent Framework, which allows you to build multi-agent systems and communicate with AWS services. He emphasized that the framework is stable and ready to use in applications.
○ Agent Building: Shawn explained that the semantic agent framework allows you to build multi-agent systems using semantic abstractions. This allows you to create agents that can communicate with each other and with external services.
○ AWS Integration: The framework supports communication with AWS services, allowing agents to interact with various cloud services to perform complex tasks.
○ Framework Stability: Shawn stated that the framework is stable and ready to be used in applications, encouraging attendees to test it and provide feedback.
• Agent Invocation Methods: Weslie explained new agent invocation methods, which allow you to create different types of agents and communicate with them using the same methods. He mentioned the use of threads to manage conversations.
○ New Methods: Weslie explained new agent invocation methods, which allow you to create different types of agents and invoke them using a uniform set of methods.
○ Thread Management: Weslie explained that threads are used to manage conversations, allowing for more efficient management of interactions between agents and users.
○ Method Uniformity: Weslie highlighted that new invocation methods allow for abstracting the differences between various types of agents, making them easier to integrate and manage.
• Memory Components for Threads: Weslie talked about adding memory components to threads, which allow for storing information from messages and building additional functionality in the future.
• Examples of Using Agents: Raj asked for examples of using different types of agents. Shawn responded that there are examples available that show how to use agents from different services.
• Collaboration with Microsoft: Luis talked about collaborating with Microsoft to integrate AI and semantic extensions. He explained that the goal is to unify abstractions and improve interoperability.
○ AI Integration: Luis discussed the collaboration with Microsoft to integrate AI and semantic extensions, with the goal of unifying abstractions and improving interoperability between different systems.
○ Unification Goal: Luis explained that the goal is to unify abstractions to allow better integration and interoperability between AI extensions and semantic systems.
○ Ongoing Collaboration: Luis highlighted that the collaboration with Microsoft is ongoing and aims to improve the user and developer experience, making it easier to use AI and semantic technologies.
• Agent Feedback: Rasmus provided feedback on agent invocation methods, suggesting changes to improve usability. Weslie responded that the feedback is useful and that changes could be made.
○ Rasmus Feedback: Rasmus suggested changes to agent invocation methods to improve usability, such as the ability to not include message content in invocations.
○ Weslie's response: Weslie responded that Rasmus' feedback is helpful and that changes could be made to improve the usability of agent invocation methods.
• Semantic Plugin Questions: Hon raised questions about semantic plugins and their advertising. Sergey responded by explaining how plugins work and how function names are handled.
○ Hon's questions: Hon asked for clarification on semantic plugins, specifically how function names are advertised and handled.
○ Sergey's response: Sergey explained that plugin function names can be changed to avoid conflicts and that the system automatically handles these changes to ensure