Overview

Configures Microsoft Semantic Kernel with AI services, creates semantic and native functions (plugins), and wires planners, memory, and telemetry to orchestrate AI workflows. It covers plugin registration, planner selection, and automatic function calling to coordinate agents like function-calling-agent and plan-and-execute-agent.

How This Skill Works

It initializes a Kernel as the central orchestrator, loads plugins as function collections, and selects a planner (Handlebars, Stepwise, or Function Calling Stepwise) to turn goals into actions. It configures AI service connectors (OpenAI or Azure), memory backends, and logging, then builds kernel function chains with automatic function calling.

When to Use It

• Setting up a function-calling-agent that coordinates multiple plugins
• Plan-and-execute-agent workflows needing Stepwise or Handlebars planning
• Integrating AI services via OpenAI or Azure connectors
• Need persistent context with a memory backend between steps
• Want automatic function calling to chain actions without manual prompts

Quick Start

1. Step 1: Install semantic-kernel and register your plugins (functions) in the Kernel
2. Step 2: Choose a planner (Handlebars, Stepwise, or Function Calling Stepwise) and configure AI service connectors
3. Step 3: Run a sample agent (function-calling-agent or plan-and-execute-agent) to verify orchestration and logging

Best Practices

• Write clear, descriptive function interfaces for plugins
• Choose the planner type that matches task complexity (Handlebars for UI-like plans, Stepwise for stepwise logic)
• Organize plugins into cohesive collections with stable API surfaces
• Implement robust error handling and fallback strategies
• Enable logging and telemetry to monitor kernel activity and function chains

Example Use Cases

• A customer-support agent that routes queries through plugins and stores context in memory
• A data enrichment workflow that plans steps and calls external services via semantic-kernel
• An automation agent that plans tasks and executes them with sequential function calls
• Using a Handlebars planner to map goals to conditional steps
• A memory-backed conversation agent that maintains context across interactions

Frequently Asked Questions