How to beginner · 3 min read

How to use Semantic Kernel function calling

Q: How to use Semantic Kernel function calling

Use semantic_kernel to define and register functions as skills, then invoke them via the kernel instance. Functions are created with kernel.create_semantic_function() and called with kernel.run() or function.invoke().

Quick answer

Use semantic_kernel to define and register functions as skills, then invoke them via the kernel instance. Functions are created with kernel.create_semantic_function() and called with kernel.run() or function.invoke().

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install semantic-kernel openai

Setup

Install the semantic-kernel Python package and set your OpenAI API key as an environment variable.

bash

pip install semantic-kernel openai

Step by step

This example shows how to create a Semantic Kernel instance, register an OpenAI chat completion service, define a semantic function, and call it.

python

import os
import semantic_kernel as sk
from semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion

# Initialize kernel
kernel = sk.Kernel()

# Add OpenAI chat completion service
kernel.add_service(OpenAIChatCompletion(
    service_id="chat",
    api_key=os.environ["OPENAI_API_KEY"],
    ai_model_id="gpt-4o-mini"
))

# Define a semantic function with a prompt template
prompt_template = """Write a friendly greeting for the user named {{name}}."""

semantic_function = kernel.create_semantic_function(
    prompt_template=prompt_template,
    skill_name="GreetingSkill",
    function_name="SayHello"
)

# Invoke the function
result = kernel.run(
    semantic_function,
    input_vars={"name": "Alice"}
)

print("Output:", result)

output

Output: Hello Alice! Hope you're having a great day!

Common variations

Use function.invoke() directly instead of kernel.run().
Switch to a different OpenAI model by changing ai_model_id.
Use async calls with await function.invoke_async() in async contexts.

python

import asyncio

async def async_example():
    result = await semantic_function.invoke_async({"name": "Bob"})
    print("Async output:", result)

asyncio.run(async_example())

output

Async output: Hello Bob! Hope you're having a great day!

Troubleshooting

If you see authentication errors, verify your OPENAI_API_KEY environment variable is set correctly.
If the function returns empty or unexpected output, check your prompt template syntax and variable names.
For model errors, confirm the ai_model_id matches a valid OpenAI model like gpt-4o-mini.

✅

Key Takeaways

Use kernel.create_semantic_function() to define reusable AI functions with prompt templates.
Invoke functions via kernel.run() or function.invoke() for flexible calling patterns.
Register AI services like OpenAI chat completions with kernel.add_service() before function creation.

Verified 2026-04 · gpt-4o-mini

Verify ↗