Microsoft is sharing its roadmap for Semantic Kernel for features that will be implemented prior to the Microsoft Build conference in a few months.
Semantic Kernel is Microsoft’s tool for adding and integrating LLM technology into existing applications.
According to the company, the roadmap groups updates into three main categories: V1.0 parity across languages, additional connectors, and agents.
The company explained that with the .NET v1.0 release, it promised to not introduce more breaking changes to non-experimental features. For its Python and Java libraries, it plans to reach Beta or Release Candidate status by March, and will release v1.0 by Microsoft Build.
Python and Java v1.0 will include many of the improvements in .NET v1.0, such as automatic function calling, events, YAML prompt files, and Handlebars templates.
“With the YAML prompt files, you’ll be able to create prompt and agent assets in Python and then reshare it with .NET and Java developers,” Matthew Bolaños, principal program manager for Semantic Kernel at Microsoft, wrote in a blog post.
Next, the company is adding new connectors to new services, such as Gemini, Llama-2, Phi-2, Mistral, and Claude, deployed on Hugging Face, Azure AI, Google AI, Bedrock, and locally. Its goal is to be able to support multi-modal experiences by adding the ability to connect to models that support audio, images, video, and documents.
Microsoft also is making improvements to its experimental memory connectors to make them less opinionated and easier to use.
And finally, the company is working on updating its agent interface so that it can support agents built with any model. Currently, its interface can only use the OpenAI Assistants API.