Presentation Generator - Optimized Version

An AI-powered presentation generator that creates professional Quarkdown presentations through a structured 4-step pipeline.

Overview

This system generates presentations by following a systematic approach:

1. Web Research & Resume: Gather comprehensive information about the topic
2. Image Collection: Find and integrate relevant images with descriptions
3. Markdown Presentation: Structure content into a presentation format
4. Quarkdown Adaptation: Convert to rich Quarkdown format for compilation

Project Structure

Core Files

• config.yml - Main configuration file (YAML-based)
• llm_optimized.py - Optimized LLM module with 4-step pipeline
• slides_pipeline_optimized.py - Main pipeline orchestrator
• slides_response_optimized.py - Enhanced response models
• generate_quarkdown.py - Quarkdown compilation utilities

Template Files

• templates/step1_research.md - Research step prompt template
• templates/step2_images.md - Image collection step prompt template
• templates/step3_markdown.md - Markdown formatting step prompt template
• templates/step4_quarkdown.md - Quarkdown adaptation step prompt template

Legacy Files (for backward compatibility)

• llm.py - Original LLM implementation
• slides_pipeline.py - Original pipeline
• slides_response.py - Original response models
• prompt.md - Original monolithic prompt

Configuration

The system is configured via config.yml:

presentation:
  llm:
    model: "openai:gpt-4.1-nano"
    max_results_per_search: 5
    image_search_limit: 5
  
  steps:
    step1_research:
      name: "Web Research & Resume"
      input: ["user_prompt"]
      output: "research_resume"
      tools: ["duckduckgo_search_tool"]
    # ... other steps

Usage

Quick Start

from slides_pipeline_optimized import prompt_to_presentation_pipeline

# Generate a presentation
prompt_to_presentation_pipeline("Create a presentation about artificial intelligence")

Advanced Usage

from llm_optimized import PresentationPipeline

# Initialize pipeline with custom config
pipeline = PresentationPipeline("custom_config.yml")

# Generate presentation
document = pipeline.generate_presentation("Your topic here")

Pipeline Steps

Step 1: Web Research & Resume

• Input: User prompt
• Output: Structured research resume
• Tools: DuckDuckGo search
• Purpose: Gather comprehensive information about the topic

Step 2: Image Collection

• Input: User prompt + Research resume
• Output: Research resume enhanced with images
• Tools: Image search (DuckDuckGo)
• Purpose: Find and integrate relevant visual content

Step 3: Markdown Presentation

• Input: User prompt + Research with images
• Output: Markdown-formatted presentation
• Tools: None (formatting only)
• Purpose: Structure content into presentation slides

Step 4: Quarkdown Adaptation

• Input: User prompt + Markdown presentation
• Output: Document (Quarkdown format)
• Tools: None (conversion only)
• Purpose: Convert to rich Quarkdown with advanced features

Key Features

Modular Design

• Each step has its own specialized prompt template
• Tools are filtered per step (only relevant tools available)
• Clear input/output definitions for each step

Configurable

• YAML-based configuration
• Customizable prompts via template files
• Adjustable parameters for each step

Step Isolation

• Each step receives only the necessary input
• Clear data flow between steps
• Independent execution of each phase

Enhanced Error Handling

• Structured error reporting
• Step-by-step execution tracking
• Graceful failure handling

Installation

1. Install dependencies:

pip install -r requirements.txt

2. Set up environment variables in .env:

OPENAI_API_KEY=your_api_key_here

3. Configure Quarkdown path in config.yml if needed

Running the System

Streamlit

streamlit run streamlit_app.py

Programmatic

from slides_pipeline_optimized import prompt_to_presentation_pipeline

prompt_to_presentation_pipeline("Your presentation topic")

Customization

Adding New Steps

1. Add step configuration to config.yml
2. Create prompt template in templates/
3. Update pipeline logic in llm_optimized.py

Modifying Prompts

Edit the corresponding template file in templates/ directory.

Changing Tools

Update the tools list in the step configuration in config.yml.

Output

The system generates:

• Quarkdown presentation file (.qmd)
• Compiled presentation (HTML/PDF via Quarkdown)
• Intermediate outputs for debugging