Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AutoGen Studio: A No-Code Developer Tool for Building and Debugging Multi-Agent Systems

Victor Dibia, Jingya Chen, Gagan Bansal, Suff Syed, Adam Fourney, Erkang Zhu, Chi Wang, Saleema Amershi

TL;DR

AutoGen Studio tackles the complexity of building multi-agent systems by providing a no-code tool with a drag-and-drop UI and declarative JSON specifications. It pairs a two-component system (frontend UI and backend API) with features for building, testing, profiling, deploying, and sharing agent workflows, all under an open-source umbrella. The work contributes design patterns for define-and-compose workflows, robust debugging and sensemaking tools, and a template gallery to accelerate development. By lowering entry barriers and enabling rapid prototyping and deployment, AutoGen Studio has the potential to accelerate research and practical adoption of autonomous multi-agent applications.

Abstract

Multi-agent systems, where multiple agents (generative AI models + tools) collaborate, are emerging as an effective pattern for solving long-running, complex tasks in numerous domains. However, specifying their parameters (such as models, tools, and orchestration mechanisms etc,.) and debugging them remains challenging for most developers. To address this challenge, we present AUTOGEN STUDIO, a no-code developer tool for rapidly prototyping, debugging, and evaluating multi-agent workflows built upon the AUTOGEN framework. AUTOGEN STUDIO offers a web interface and a Python API for representing LLM-enabled agents using a declarative (JSON-based) specification. It provides an intuitive drag-and-drop UI for agent workflow specification, interactive evaluation and debugging of workflows, and a gallery of reusable agent components. We highlight four design principles for no-code multi-agent developer tools and contribute an open-source implementation at https://github.com/microsoft/autogen/tree/main/samples/apps/autogen-studio

AutoGen Studio: A No-Code Developer Tool for Building and Debugging Multi-Agent Systems

TL;DR

AutoGen Studio tackles the complexity of building multi-agent systems by providing a no-code tool with a drag-and-drop UI and declarative JSON specifications. It pairs a two-component system (frontend UI and backend API) with features for building, testing, profiling, deploying, and sharing agent workflows, all under an open-source umbrella. The work contributes design patterns for define-and-compose workflows, robust debugging and sensemaking tools, and a template gallery to accelerate development. By lowering entry barriers and enabling rapid prototyping and deployment, AutoGen Studio has the potential to accelerate research and practical adoption of autonomous multi-agent applications.

Abstract

Multi-agent systems, where multiple agents (generative AI models + tools) collaborate, are emerging as an effective pattern for solving long-running, complex tasks in numerous domains. However, specifying their parameters (such as models, tools, and orchestration mechanisms etc,.) and debugging them remains challenging for most developers. To address this challenge, we present AUTOGEN STUDIO, a no-code developer tool for rapidly prototyping, debugging, and evaluating multi-agent workflows built upon the AUTOGEN framework. AUTOGEN STUDIO offers a web interface and a Python API for representing LLM-enabled agents using a declarative (JSON-based) specification. It provides an intuitive drag-and-drop UI for agent workflow specification, interactive evaluation and debugging of workflows, and a gallery of reusable agent components. We highlight four design principles for no-code multi-agent developer tools and contribute an open-source implementation at https://github.com/microsoft/autogen/tree/main/samples/apps/autogen-studio
Paper Structure (25 sections, 5 figures)

This paper contains 25 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Agents ( llms + Tools)
  4. multi-agent Frameworks
  5. Design Goals
  6. System Design
  7. User Interface
  8. Building Workflows
  9. Testing and Debugging Workflows
  10. Deploying Workflows
  11. Template Gallery
  12. Backend API - Web, Python, and Command Line
  13. Usage and Evaluation
  14. Emerging Design Patterns and Research Directions
  15. Define-and-Compose Workflows
  16. ...and 10 more sections

Figures (5)

  • Figure 1: AutoGen Studio provides a drag-n-drop UI where models, skills/tools, memory components can be defined, attached to agents and agents attached to workflows.
  • Figure 2: AutoGen Studio provides a backend api (web, python, cli) and a UI which implements a playground (shown), build and gallery view. In the playground view, users can run tasks in a session based on a workflow. Users can also observe actions taken by agents, reviewing agent messages and metrics based on a profiler module.
  • Figure 3: AutoGen Studio can be installed from PyPI (pip) and the UI launched from the command line.
  • Figure 4: Workflows can be imported in python apps.
  • Figure 5: Plot of GitHub issues ($n=8$ clusters) from the AutoGen Studio repo. User feedback ranged from support with workflow authoring tools (e.g., the ability configure and test models) to general installation.