MultiAgentPPT is an intelligent presentation generation system that automates the creation of editable PPTs from a given topic. It targets users who need to quickly produce structured and informative presentations, leveraging a multi-agent architecture for efficiency and quality.

How It Works

The system employs a multi-agent workflow orchestrated by A2A, MCP, and ADK frameworks. It begins with an Outline Agent to generate a topic outline, followed by a Topic Split Agent. Multiple Research Agents then work in parallel to gather information on each sub-topic, utilizing RAG for knowledge retrieval. Finally, a Summary Agent, through a loop of PPT generation and checking agents, compiles the research into slide content, which is streamed back to the frontend. This parallel and iterative approach enhances efficiency and allows for handling LLM output limitations.

Quick Start & Requirements

• Backend:
  • Create and activate a Conda environment (Python 3.12 recommended).
  • Install backend dependencies: pip install -r requirements.txt in the backend directory.
  • Configure environment variables by copying .env from env_template in each backend module directory.
  • Start backend services (e.g., python main_api.py for slide_agent on port 10011).
• Frontend:
  • Run PostgreSQL via Docker: docker run --name postgresdb -p 5432:5432 -e POSTGRES_USER=postgres -e POSTGRES_PASSWORD=welcome -d postgres.
  • Configure DATABASE_URL and agent URLs in the frontend's .env file.
  • Install frontend dependencies: pnpm install.
  • Push database models: pnpm db:push.
  • Start frontend: npm run dev.
• Access: http://localhost:3000/presentation

Highlighted Details

• Supports parallel research by multiple agents for improved efficiency.
• Features real-time streaming of generated PPT content to the frontend.
• Includes a PPTChecker Agent for quality assurance of each slide.
• Allows for local model adaptation and metadata transfer between frontend and agents.

Maintenance & Community

The project is maintained by johnson7788. Contact information for author WeChat is provided for Q&A.

Licensing & Compatibility

The README does not explicitly state the license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The system is demonstrated with example data ("Overview of Electric Vehicle Development"); configuring it for other topics requires setting up corresponding agents and data sources. The frontend integration for downloadable PPTX is listed as a "to-do" item.