Skip to content
Nappai Documentation

Slide Generator Agent

Slide Generator Agent

The Slide Generator Agent is an AI assistant designed to automatically convert text, data, or ideas into a slide presentation. Its main function is to take input content and structure it into a presentation format (.pptx or Markdown for Marp), saving you the time of manually designing each slide.

Main Use Cases

Use the Slide Generator Agent to:

  • Create presentation drafts from a document or a summary.
  • Transform meeting notes into a structured presentation.
  • Automatically generate visual reports from data.
  • Quickly create educational or marketing content.

Configuration (Common Tab)

These are the main parameters to control the generation of your presentation.

ParameterTechnical DescriptionPurpose in the Flow
User InstructionsThe template that formats the user input.This is where you define the request structure. The {input} variable will be replaced with the text provided in the Input field.
InputThe base content for the presentation.This is where you paste the text, data, summary, or topic you want the agent to use to generate the slides.
Output FormatThe file format in which the presentation will be generated.You can choose between pptx (for Microsoft PowerPoint) or text-based formats like marp (Markdown).
Marp ThemeThe visual theme applied to the presentation if you choose a marp format.Allows customizing the appearance of your slides (colors, fonts, etc.) using Marp’s predefined themes.
ToolsConnection for external tools the agent can use.Allows the agent to, for example, search for additional info on the web or query a database to enrich the presentation content.
ModelConnection to the Language Model (LLM) that will generate the slide content.It is the agent’s “creative brain.” Connect a component like Gemini or OpenAI here.

Configuration (Advanced Tab)

The Advanced tab for this agent is very similar to the Simple Agent, allowing you to control aspects such as maximum iterations, the use of conversational memory, or streaming output. Refer to the Simple Agent documentation for a detailed breakdown of these options.

How It Works Internally

  1. Content Interpretation: The agent receives the text from the Input field.
  2. LLM Call: It sends this content to the connected Language Model (LLM), along with the instruction to structure it as a presentation.
  3. Markdown Generation: The LLM returns the content organized in a special Markdown format compatible with Marp, a tool for creating slides from text.
  4. Conversion to Final Format: If you selected pptx, NappAI internally uses a tool to convert that Markdown into a PowerPoint file. If you chose marp, it delivers the text file ready for use.

Usage Example: Creating a Presentation from a Summary

Goal: Transform an article summary into a 5-slide presentation.

  1. Configure Inputs:

    • Input: Paste the text of the article summary.
    • User Instructions: Modify it to be more specific: Create a 5-slide presentation based on the following text: {input}.
    • Output Format: Select pptx.
    • Model: Connect a language model like gpt-4o-mini.

  2. Run the Flow: The agent will process the text and generate a .pptx file.

  3. Review the Output:

    • The main output of the component will be the binary file of the presentation, which you can save or send to another component (for example, Gmail Message to send it by email).

Tips and Best Practices

  • Be Specific in Instructions: To get better results, guide the agent. Instead of just pasting text, in the Input or User Instructions field, you can give directives like: “Create a presentation with a title slide, three content slides, and a conclusion slide on the following topic: …”.
  • Install Marp CLI for Previews: If you work extensively with the marp format, having the marp-cli tool installed on your computer will allow you to convert and view presentations locally.
  • Use Tools to Enrich: Connect a Web Search to the Tools input and ask the agent: “Search for information about NappAI and create a presentation.” The agent will first research and then use that information to generate the slides.