johnson7788/MultiAgentPPT

What it solves

MultiAgentPPT automates the creation of high-quality, editable presentation slides from a single topic input. It solves the problem of manual content research and slide structuring by using a coordinated system of AI agents to handle everything from outlining to deep research and final content generation.

How it works

The system employs a multi-agent pipeline to transform a topic into a presentation:

1. Outline Generation: An agent creates an initial structure based on user requirements.
2. Topic Splitting: A specialized agent breaks the outline into smaller, manageable themes.
3. Parallel Research: Multiple Research Agents work concurrently to perform deep dives into each theme using RAG (Retrieval-Augmented Generation) and MCP (Model Context Protocol) tools.
4. Synthesis: A Summary Agent aggregates the research findings into slide content.
5. Iterative Generation: A loop-based PPT Agent generates slides page-by-page, while a PPTChecker Agent reviews each page for quality, retrying up to three times if necessary.

Who it’s for

It is designed for users who need to quickly generate professional, research-backed presentation content without manually drafting every slide or conducting separate research for each section.

Highlights

• Concurrent Agent Architecture: Uses parallel agents to increase efficiency and accuracy of content generation.
• Streaming Output: Provides real-time streaming of the outline and slide content to the frontend for a better user experience.
• Quality Control: Includes a dedicated checker agent to validate the quality of each generated slide.
• RAG Integration: Leverages external knowledge base retrieval to ensure content is grounded in actual data.
• Export Capability: Supports downloading the generated content as .pptx files via python-pptx.