microsoft/semantic-kernel

Features

• Agent Orchestration Frameworks - Building intelligent systems that coordinate multiple language models and external tools to complete complex, multi-step business tasks.
• AI Orchestration Frameworks - A development platform that integrates large language models with existing code to automate complex tasks and business workflows.
• Model Abstraction Layers - A standardized interface layer decouples core orchestration logic from specific large language model providers and their proprietary API protocols.
• Prompt Engineering Tools - A text-processing system that dynamically injects runtime variables and context into structured prompts before dispatching them to an external model.
• Agentic Runtimes - A runtime environment for building autonomous agents that execute multi-step processes by chaining together specialized functions and tools.
• Function Orchestrators - A modular system where discrete functional units are dynamically discovered and executed by a central planner to satisfy complex user requests.
• Task Planners - An automated reasoning engine that breaks high-level goals into sequences of executable steps by analyzing available function signatures and their metadata.
• Vector Retrieval Abstractions - A unified interface layer that maps diverse vector databases and storage providers into a consistent retrieval mechanism for context-aware AI operations.
• Enterprise AI Integration Tools - Connecting generative artificial intelligence models to proprietary corporate data and existing software infrastructure while maintaining strict security and compliance standards.
• Language-Agnostic Connectors - A standardized interface layer that decouples the core orchestration logic from specific large language model providers and their proprietary API protocols.
• LLM Integration Layers - Providing a standardized software interface that connects generative artificial intelligence models to traditional application logic and data sources.
• Agent Frameworks - Integrates AI models with enterprise-grade security and plugin support.
• AI Agent Frameworks - Integration of LLMs into existing applications.
• AI Frameworks and SDKs - SDK for integrating LLMs into C#, Python, and Java.
• AI & Machine Learning - SDK for integrating LLMs into conventional programming languages.
• Application Development - Library for prompt templating, function chaining, and planning.
• Application Frameworks - SDK to integrate LLM technology into applications.
• Language Model Development - SDK for integrating LLMs with conventional programming code.
• LLM Frameworks and Libraries - Integrates LLM technology into applications via composable components.
• RAG Frameworks - SDK for developing generative AI applications with integrated LLM support.
• AI and Machine Learning - SDK for orchestrating AI models and building intelligent agents.
• Agentic AI - Listed in the “Agentic AI” section of the The Incredible Pytorch awesome list.
• Workflow Automation Engines - Streamlining repetitive business processes by chaining together automated reasoning steps and external service calls into a single cohesive pipeline.
• Middleware Pipelines - A chain of interceptors that process prompts and responses to inject cross-cutting concerns like logging, telemetry, and safety filtering during execution.
• Plugin Architectures - Add custom functionality by plugging in external components that integrate seamlessly with existing logic to meet unique project requirements without modifying the core structure of your application.
• Modular AI Components - Creating flexible and reusable AI components that can be easily swapped, tested, and scaled within larger software applications.

Star 历史