ai-garage

System Design Agent (using Autogen)

Overview

Inspired by Alex Xu’s ByteByteGo edition on “How to ace System Design Interviews”, this System Design Agent is an advanced multi-agent tool designed to assist software architects and engineers in creating robust, scalable, and efficient design documents for any proposed sopftware system.

It leverages the sequential chat pattern in Autogen for a multi-agent system to break down complex design tasks into manageable components, ensuring a thorough and well-structured approach to system architecture. The agents collaborate to clarify requirements, estimate capacity, design high-level architecture, choose optimal database types, and design interfaces, among other tasks.

Agents

The following table describes the various agents that are part of this multi-agent system:

Agent	Description
requirement_clarification_agent	Ensures a complete understanding of system requirements by asking critical questions.
capacity_estimation_agent	Estimates system resource needs based on requirements using back-of-the-envelope calculations.
high_level_design_agent	Designs & visualizes the high-level architecture, breaking it into critical components.
database_design_agent	Models data and designs efficient data storage systems.
interface_design_agent	Designs interfaces between system components, choosing appropriate communication methods.
scalability_and_performance_agent	Ensures the system can scale and maintain efficiency under load.
reliability_and_resilience_agent	Designs fault-tolerant and resilient systems, ensuring high availability and reliability.
writer_agent	Writes well-structured, formatted, and engaging software design documentation in markdown format.

Features

• Multi-Agent System: The project utilizes multiple specialized agents, each focusing on a specific aspect of system design.
• CLI Interface: A user-friendly command line interface for interacting with the agents and executing tasks.
• Human-in-the-loop for Requirement Specification: Given an initial problem statement, the system design agent confirms your actual requirements & specifications by asking relevant questions, and then proceeds for automated design.
• Visualization of High-Level ARchitecture & Database Schemas: Dedicated agents produce PLantUML syntax to visualize teh various components of the high-level architecture, and also the data
• Design Document produced at the End: Based on the agentic responses, a detailed, well-formatted design document is produced in Markdown, available in the sample_systems_designed folder.

Setup & Usage Instructions

Prerequisites

• Python 3.10 or higher
• Poetry [Installation Instructions]
• Setup your GPT Model on Azure OpenAI [Reference Video]

Installation

1. Clone the repository & get into the project directory

   git clone https://github.com/tezansahu/ai-garage.git
   cd ai-garage/diagram-generator
   

2. Install dependencies using Poetry:

   poetry shell
   poetry install #  (installs Autogen & other necessary packages) 
   

3. Set up your environment variables in .env file (add the necessary environment variables as specified in .env.example)

Usage

1. Run the CLI

    python app.py
   

2. Follow the Prompts
   • Enter the problem statement when prompted.
   • The system will guide you through the process of designing the system, asking relevant questions and providing detailed outputs.

Conclusion

The System Design Agent Project is a powerful tool for software architects and engineers, providing a structured and efficient approach to system design. By leveraging a multi-agent system and a user-friendly CLI, this project ensures that all aspects of system design are thoroughly addressed, resulting in robust and scalable architectures.

This site is open source. Improve this page.