The Microsoft Semantic Kernel team announced the project's fall 2023 release plan. "Our focus is to deliver key developments and enhancements in three pillar areas: open source and trustworthiness, reliability and performance, and integration of the latest AI innovations."
The specific content includes:
- Announced adoption of the OpenAI Plugin open standard, allowing users to create plugins that work across OpenAI, Semantic Kernel, and Microsoft platforms. The team will be working on the scalability and reliability of the core kernel and planner.
- There will be work on improving reliability and performance by improving Planners , making them more efficient and able to easily handle global-scale deployments. The team will provide cold storage plans to ensure consistency and introduce features such as a dynamic planner that automatically discovers plugins. A Planner is a function that takes user queries and returns a plan for how to complete the request. For example , if you have task and calendar event plugins, Planner can combine them to create workflows. Like "remind me to go to the store for milk" or "remind me to call mom tomorrow" without having to explicitly code those scenarios.
- Integrate Semantic Kernel with vector databases such as Pine Cone, Redis, Weaviate, Chroma, and Azure Cognitive Search and Services . And plans to add document chunking service and improve VS Code extension.
In a blog post, John Maeda, Microsoft's vice president of design and artificial intelligence, said telemetry and AI safety will be high on the agenda. With end-to-end telemetry, developers gain insight into goal-oriented AI program creation, token usage, and errors. The integration of Azure Content Safety hooks provides a simplified approach to ensuring AI safety.
In general, features expected to be added include:
- plugin test
- dynamic planner
- end-to-end telemetry
- streaming support
- Enhanced VS Code extension
"Semantic Kernel's Fall 2023 release plan emphasizes open collaboration, performance improvements, and integration of cutting-edge AI innovations. Together, we are building a robust and reliable AI development platform."