I Built an AI That Writes & Runs Code 🤯 | AutoGen Code Executor (Python + Groq)

This tutorial demonstrates building an AI agent using AutoGen and Groq that can autonomously write and execute Python code. The project leverages AutoGen's multi-agent framework to create agents capable of code generation and execution, powered by Groq's fast LLM inference. The implementation showcases practical AI automation for code-based tasks using Python.

Key Points

• •AutoGen enables multi-agent collaboration where agents can write, review, and execute code autonomously
• •Groq provides fast LLM inference for real-time code generation and execution workflows
• •Code executor agents can validate and run Python scripts within a controlled environment
• •Multi-agent architecture allows separation of concerns: one agent writes code, another executes and validates
• •Integration of LLM capabilities with local code execution creates a powerful automation pipeline
• •Groq's speed advantage is critical for interactive code generation and debugging cycles
• •AutoGen handles agent communication and orchestration automatically
• •Python environment setup and dependency management are essential for safe code execution

Workflow Diagram

Concepts

Artifacts (2)

# Basic AutoGen setup with Groq for code execution
from autogen import AssistantAgent, UserProxyAgent

# Initialize agents
code_writer = AssistantAgent(
    name="CodeWriter",
    system_message="You are an expert Python programmer. Write clean, efficient code.",
    llm_config={"model": "groq", "api_key": "your_groq_api_key"}
)

code_executor = UserProxyAgent(
    name="CodeExecutor",
    human_input_mode="NEVER",
    code_execution_config={"work_dir": "./code_output", "use_docker": False}
)

# Start conversation
code_executor.initiate_chat(
    code_writer,
    message="Write a Python script that calculates fibonacci numbers"
)

{
  "llm_config": {
    "config_list": [
      {
        "model": "mixtral-8x7b-32768",
        "api_key": "${GROQ_API_KEY}",
        "api_type": "groq"
      }
    ],
    "temperature": 0.7,
    "timeout": 120
  },
  "code_execution": {
    "work_dir": "./code_output",
    "use_docker": false,
    "timeout": 60
  }
}