Semantic Kernel Code Execution Example

This repository contains examples of using Semantic Kernel to execute Python code either locally or in a sandboxed environment. The examples demonstrate how to set up and use ChatCompletionAgent to process user input, generate Python code, and execute it safely.

Features

• Azure OpenAI Integration: Uses Azure OpenAI services for code generation.
• Safe Code Execution: Executes Python code in a restricted environment to ensure safety.
• Logging: Provides detailed logging for debugging and monitoring.
• Environment Configuration: Uses environment variables for configuration.
• Flexible Execution: Can be executed locally or in the Azure Container Apps Dynamic Sessions Code Interpreter.
• Agent Group Chat: Demonstrates how to create a group chat with multiple agents working together.

Getting Started

Prerequisites

• Python 3.12
• Azure OpenAI account

Installation

1. Clone the repository:

   git clone https://github.com/john-carroll-sw/semantic-kernel-agent-code-execution.git
   cd semantic-kernel-agent-code-execution

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install the required packages:

   pip install -r requirements.txt

4. Set up your environment variables in a .env file. You can use the provided .env.sample as a template:

   AZURE_OPENAI_ENDPOINT=https://your-openai-endpoint.openai.azure.com/
   AZURE_OPENAI_API_KEY="your-api-key"
   AZURE_OPENAI_API_VERSION=2024-08-01-preview
   AZURE_OPENAI_DEPLOYMENT=your-deployment-name
   AZURE_CODE_INTERPRETER_POOL_ENDPOINT=your-pool-endpoint

   • AZURE_OPENAI_ENDPOINT: The endpoint for your Azure OpenAI service.
   • AZURE_OPENAI_API_KEY: Your Azure OpenAI API key.
   • AZURE_OPENAI_API_VERSION: The API version to use for Azure OpenAI.
   • AZURE_OPENAI_DEPLOYMENT: The deployment name for your Azure OpenAI service.
   • AZURE_CODE_INTERPRETER_POOL_ENDPOINT: The endpoint for managing the Azure Container Apps Dynamic Sessions Code Interpreter. Only add this if you plan on using the ACA Dynamic Sessions Code Interpreter which uses the SessionsPythonTool.

Usage

By default, the generated code is set to run locally. If you want to run the code in a sandboxed environment, make sure to configure the AZURE_CODE_INTERPRETER_POOL_ENDPOINT and set USE_CODE_INTERPRETER_SESSIONS_TOOL to True in the respective scripts.

Code Execution Example

Run the script:

python code_execution_example.py

Enter your message when prompted, and the agent will process it, generate Python code, and execute it. This example demonstrates how to execute code either locally or in a sandboxed environment.

Agent Group Code Execution Example

Run the script:

python agent_group_code_execution.py

Enter your message when prompted, and the agents will work together to generate and execute Python code. This example also supports executing code either locally or in a sandboxed environment.

Agent Group Writing Example

Run the script:

python agent_group_writing_example.py

Enter your message when prompted, and the agents will work together to review and rewrite the content.

Example Questions for Code Interpreter

For examples of good questions or prompts to use with a code interpreter, refer to the code_interpreter_questions.md file.

License

This project is licensed under the MIT License.