/ollama-presentation

/ollama-presentation

Create a PowerPoint presentation with AI-generated visual assets. You orchestrate two capabilities:

1. Image generation — via src/generate_image.py and the ollama-image-expert agent for custom images, icons, diagrams, and patterns
2. Slide creation — via pptxgenjs (the pptx skill's creation engine)

Consult the ollama-image-expert agent for all image generation knowledge, including the "Images for Presentations" section specifically.

Parse Arguments

Parse $ARGUMENTS for:

• Topic/description: The quoted text describing what the presentation is about
• --slides N: Approximate number of slides (default: 8-12)
• --output PATH: Output path (default: output/presentation-YYYYMMDD-HHMMSS.pptx)
• --style STYLE: Visual style — corporate, creative, technical, minimal (default: infer from topic)
• --prompt-file PATH: Read a detailed brief from a file
• --no-images: Skip image generation (text and shapes only, much faster)

If no topic and no --prompt-file, stop and ask what the presentation is about.

Step 1: Plan the Presentation

Before generating anything, plan the full deck. Create a slide plan with:

For each slide:

• Slide number and title
• Content summary (key points, data, or message)
• Layout type: title-slide, two-column, icon-grid, stat-callout, image-feature, diagram, section-divider, closing
• Visual asset needed: what image/icon/diagram/pattern would enhance this slide
• Asset generation approach: which skill and prompt to use

Visual asset strategy by layout:

Layout	Visual Asset	Generation Approach
Title slide	Background pattern or hero image	/ollama-pattern for texture background, or /ollama-image for hero
Two-column	Content illustration	/ollama-image with topic-specific prompt
Icon grid	3-4 custom icons	/ollama-icon --style flat for each concept
Stat callout	Background accent or pattern	/ollama-pattern or use shapes only
Image feature	Full or half-bleed image	/ollama-image with composition-aware prompt
Diagram	Technical diagram	/ollama-diagram --type flowchart/architecture
Section divider	Subtle pattern or gradient image	/ollama-pattern --style abstract
Closing	Echo title slide background	Reuse title slide assets

Not every slide needs a generated image. Charts, shapes, and icons from react-icons are faster and sometimes better. Use generated images where they add genuine value — hero shots, illustrations of concepts, custom diagrams, branded patterns.

Step 2: Choose Color Palette and Typography

Pick a palette that matches the topic (don't default to blue). Reference the pptx skill's palette table or create a custom one. Define:

• Primary color (60-70% weight)
• Secondary color
• Accent color
• Background colors (dark for title/closing, light for content)
• Text colors (primary, muted)

Pick a font pairing — header font with personality, clean body font.

Locate Plugin

Before running any Python scripts, discover the plugin root:

PLUGIN_ROOT=$(python3 -c "
from pathlib import Path
import sys, os

if os.environ.get('JACK_TAR_OLLAMA_ROOT'):
    print(os.environ['JACK_TAR_OLLAMA_ROOT']); sys.exit()

home = Path.home()
for base in [home / '.claude' / 'plugins' / 'cache']:
    for p in base.rglob('jack-tar-ollama/.claude-plugin/plugin.json'):
        print(str(p.parent.parent)); sys.exit()

dev = Path.cwd() / 'plugins' / 'jack-tar-ollama'
if dev.exists():
    print(str(dev)); sys.exit()

print('NOT_FOUND')
" 2>/dev/null)
if [ -z "$PLUGIN_ROOT" ] || [ "$PLUGIN_ROOT" = "NOT_FOUND" ]; then
  echo "ERROR: jack-tar-ollama plugin not found. Set JACK_TAR_OLLAMA_ROOT or install the plugin."
  exit 1
fi

Step 3: Generate Visual Assets

Generate all needed images BEFORE building slides. Store them in a temporary directory:

mkdir -p /tmp/presentation-assets

For each planned visual asset, run the appropriate generation:

For content images:

python3 "$PLUGIN_ROOT/src/generate_image.py" --prompt "PROMPT" --model "MODEL" --output /tmp/presentation-assets/slide-N-image.png --width WIDTH --height HEIGHT

Image sizing for slides (16:9 at 10" × 5.625"):

• Full-bleed background: 1920x1080 (or 1024x576 for faster generation)
• Half-slide image (left or right): 960x1080 (or 512x576)
• Content image in layout: 768x768 or 1024x768
• Icon: 512x512

Consult the ollama-image-expert agent for prompt strategies, especially the "Images for Presentations" section for slide-specific guidance.

For icons — if react-icons doesn't have what you need:

python3 "$PLUGIN_ROOT/src/generate_image.py" --prompt "ICON PROMPT" --model "x/flux2-klein" --output /tmp/presentation-assets/icon-N.png --width 512 --height 512 --steps 20

For diagrams:

python3 "$PLUGIN_ROOT/src/generate_image.py" --prompt "DIAGRAM PROMPT" --model "x/flux2-klein" --output /tmp/presentation-assets/diagram-N.png --width 1024 --height 768 --steps 20

For patterns/textures:

python3 "$PLUGIN_ROOT/src/generate_image.py" --prompt "PATTERN PROMPT" --model "x/z-image-turbo" --output /tmp/presentation-assets/pattern-N.png --width 1024 --height 1024 --steps 8

Review each generated image using the Read tool. If an image is poor quality or doesn't match the intent, regenerate with a refined prompt. Don't waste time on perfection — one retry is usually enough for presentations.

Step 4: Build the Presentation

Ensure pptxgenjs is installed:

which pptxgenjs || npm install -g pptxgenjs

Write a Node.js script that builds the presentation using pptxgenjs. Follow the pptx skill's creation guide (pptxgenjs.md) for API reference. Key points:

• Use slide.addImage({ path: "/tmp/presentation-assets/slide-N-image.png", ... }) for generated images
• Use slide.background = { path: "/tmp/presentation-assets/pattern-N.png" } for pattern backgrounds
• Set proper sizing with sizing: { type: 'cover', w: W, h: H } for backgrounds
• Use sizing: { type: 'contain', w: W, h: H } for content images to preserve aspect ratio

Follow pptx skill design guidelines:

• Every slide needs a visual element
• Don't repeat the same layout across slides
• Left-align body text, center only titles
• 0.5" minimum margins
• Never use # prefix on hex colors
• Don't reuse option objects (pptxgenjs mutates them)
• NEVER use accent lines under titles

Run the script:

node /tmp/presentation-build.js

Step 5: QA

Content QA:

pip install "markitdown[pptx]" 2>/dev/null
python -m markitdown OUTPUT_PATH

Check for missing content, wrong order, typos.

Visual QA: Convert to images and review:

python scripts/office/soffice.py --headless --convert-to pdf OUTPUT_PATH 2>/dev/null || soffice --headless --convert-to pdf OUTPUT_PATH
pdftoppm -jpeg -r 150 OUTPUT_PDF slide

Read the slide images with the Read tool. Check for:

• Overlapping elements
• Text overflow or cut off
• Generated images that look wrong or garbled
• Low contrast text
• Uneven spacing

Fix any issues found. Re-verify after fixes.

Step 6: Clean Up and Report

rm -rf /tmp/presentation-assets /tmp/presentation-build.js

Report:

• Path to the generated .pptx file
• Number of slides
• Number of AI-generated images used
• The color palette and fonts chosen
• Any known limitations (e.g., "text in the diagram on slide 5 may not be fully legible")

When to Skip Image Generation

Use --no-images or skip image generation when:

• The user needs a quick text-focused deck
• Ollama isn't running
• Speed matters more than visuals
• The topic is better served by charts and data than illustrations

In these cases, use shapes, react-icons, charts, and strong typography to create visual interest without generated images.