MultiAgentPPT: A PPT Auto-Generation System Integrating A2A + MCP + ADK 80

MultiAgentppt is a project that combines multi-agent systems, the A2A framework, the MCP collaboration mechanism, and Google ADK technology. Through intelligent agent collaboration and a streaming concurrency mechanism, it achieves full-process automation from inputting a topic to automatically generating a complete PPT.

System Workflow

• Outline Generation Agent: Generates a clearly structured content framework based on the topic input by the user.

• Topic Splitting Agent: Breaks down the outline into multiple sub-topic modules.

• Research Agents (multiple): Independently retrieve information for each sub-topic and generate corresponding text content.

• Summary Agent: Aggregates all research results and consolidates them into copy that can be directly used for the PPT.

Core Features

• Multi-agent division of labor: Different agents focus on a single task, improving overall efficiency and optimizing content structure.

• Parallel processing: Multiple Research Agents simultaneously conduct information retrieval and generation, reducing creation time.

• Real-time streaming output: Results are fed back instantly during content generation, eliminating waiting.

• Modular design: Supports flexible replacement or extension of agents to adapt to various scenario requirements.

• External data integration: Connects to real data sources to enhance the authority and practicality of generated content.

Advantage Summary

• Multi-agent parallel work improves content generation efficiency and accuracy;

• Supports streaming content return, optimizing the user's real-time viewing experience;

• Combines external retrieval with intelligent collaboration to ensure high-quality output of outlines and PPTs;

• Flexible system architecture facilitates the addition of new features or extension of agent modules.

Usage Interface Workflow

• The system provides complete operational interface guidance, covering the entire process from topic input to PPT generation:

• Topic Input Page: The user fills in the PPT topic they want to create (e.g., "Overview of Electric Vehicle Development").

• Outline Streaming Generation: The system returns a progressively refined outline structure in real time based on the input.

• Complete Outline Confirmation: The final complete outline is presented for the user to review and adjust.

• PPT Content Streaming Generation: After the outline is confirmed, the system generates slide content page by page and synchronizes it to the frontend.

Multi-agent Collaboration Workflow

The collaboration logic can be intuitively understood through the flowchart:

• The Outline Generation Agent creates the initial framework;

• The Topic Splitting Agent refines sub-topics;

• Multiple Research Agents conduct parallel research on each sub-topic;

• The Summary Agent integrates the materials to generate the final PPT copy;

• The generated content is streamed back to the user interface in real time.

Environment Configuration Guide

Backend Deployment (Python):

Use conda to create a virtual environment named multiagent, activate it, and then enter the backend directory to install dependencies. You need to enter the simpleOutline, simplePPT, slide_outline, and slide_agent subdirectories respectively, copy the env_template file to .env, and configure the parameters. Taking slide_agent as an example, enter the directory and run python main_api.py to start the service.

Frontend Running:

Based on the Next.js framework, enter the frontend directory to install dependencies, copy the .env file, and then execute npm run dev. Access http://localhost:3000/ through a browser to launch the application.

Project Address:https://github.com/johnson7788/MultiAgentPPT