# Generate from topic
python -m scripts.carousel_generator "GLP-1 for weight loss" --template tips_5

# Generate from JSON structure
python -m scripts.carousel_generator input.json --output ./my-carousel/

# Generate from long-form content (backward mode)
python -m scripts.carousel_generator newsletter.txt

# Generate both 4:5 and 1:1 outputs
python -m scripts.carousel_generator "Statin myths" --template myth_busting --both-ratios

"Create a carousel about statins myth-busting"
"Generate Instagram slides for GLP-1 benefits"
"Turn this newsletter into carousel slides: [paste content]"

INPUT LAYER
├── Topic only ("GLP-1 for weight loss")
├── Long-form content (newsletter, script, thread)
└── Structured JSON (slides already defined)
        ↓
CONTENT INTELLIGENCE (Claude/GPT-4o-mini)
├── Research via PubMed MCP + AstraDB RAG
├── Classify: 4A Framework (Actionable/Analytical/etc.)
├── Select template type (Tips/Stats/Story/Compare)
├── Structure into 8-10 slides with hook + CTA
└── Output: Structured JSON with slide specs
        ↓
VISUAL GENERATION (Multi-Tool Routing)
├── Standard slides → React + Puppeteer (default, high fidelity)
├── Pillow → Fallback renderer if Puppeteer unavailable
├── Infographics → Gemini API (medical accuracy)
├── Data slides → Plotly (charts, forest plots)
└── Icons → Health Icons (free, SVG, medical)
        ↓
QUALITY GATES
├── WCAG AA contrast check (4.5:1 ratio)
├── Text density check (≤15 words per slide)
├── Anti-AI voice verification
├── Brand consistency (design tokens)
└── Mobile preview generation
        ↓
OUTPUT
├── 8-10 PNG slides (1080×1350 or 1080×1080)
├── Dual ratio naming: slide_01_4x5.png, slide_01_1x1.png
├── Caption suggestions per slide
├── Alt text for accessibility
├── Hashtag recommendations
└── Quality report

HOOK → TIPS → TIPS → TIPS → TIPS → TIPS → STATS → CTA

HOOK → MYTH → MYTH → MYTH → STATS → CTA

HOOK → STORY → STORY → DATA → TIPS → QUOTE → CTA

HOOK → STATS → DATA → COMPARISON → TIPS → CTA

HOOK → STEPS → STEPS → STEPS → TIPS → CTA

python -m scripts.carousel_generator INPUT [OPTIONS]

Arguments:
  INPUT                    Topic string, JSON file, or text file

Options:
  -t, --template TEXT      Template preset (tips_5, myth_busting, etc.)
  -a, --account INTEGER    Account: 1=@heartdocshailesh, 2=@dr.shailesh.singh
  -r, --ratio TEXT         Aspect ratio: 4:5 (Instagram) or 1:1 (square)
  --both-ratios            Generate both 4:5 and 1:1 outputs
  -o, --output PATH        Output directory
  --no-ai                  Skip AI content structuring (uses curated database if available)

# Basic usage
python -m scripts.carousel_generator "5 signs of heart attack"

# Myth-busting format
python -m scripts.carousel_generator "Statin myths" --template myth_busting

# Use secondary account
python -m scripts.carousel_generator "CAC scoring" --account 2

# Square format for multi-platform
python -m scripts.carousel_generator "BP monitoring" --ratio 1:1

# Dual ratio output
python -m scripts.carousel_generator "Hypertension myths" --both-ratios

# Custom output directory
python -m scripts.carousel_generator "GLP-1" -o ./client-carousel/

from scripts.carousel_generator import CarouselGenerator
from scripts.models import CarouselConfig, AspectRatio

# Configure
config = CarouselConfig(
    account=1,
    aspect_ratio=AspectRatio.INSTAGRAM_4X5,
    max_slides=10,
    check_contrast=True,
    check_anti_ai=True
)

# Initialize generator
generator = CarouselGenerator(config)

# Generate from topic
result = generator.generate_from_topic(
    "GLP-1 for weight loss",
    template="tips_5",
    use_ai=True
)

print(f"Generated {len(result.slides)} slides to {result.output_directory}")

# Create structured input
carousel_json = {
    "topic": "Heart Health Tips",
    "slides": [
        {
            "type": "hook",
            "title": "5 Things Your Cardiologist Wishes You Knew",
            "subtitle": "Evidence-based insights"
        },
        {
            "type": "tips",
            "title": "Tip #1",
            "bullet_points": [
                "LDL target matters more than total cholesterol",
                "Below 100 mg/dL for most, below 70 for high risk"
            ]
        },
        {
            "type": "cta",
            "cta_text": "Follow for more",
            "cta_handle": "@heartdocshailesh"
        }
    ]
}

import json
with open("input.json", "w") as f:
    json.dump(carousel_json, f)

result = generator.generate_from_json(Path("input.json"))

newsletter_content = """
# The Truth About Statins

Recent meta-analyses have conclusively shown that statins reduce
cardiovascular events by 25-30%. The most common side effect,
muscle pain, occurs in only 5-10% of patients...
"""

result = generator.generate_from_longform(
    newsletter_content,
    content_type="newsletter"
)

from scripts.quality_checker import QualityChecker
from scripts.models import Carousel

checker = QualityChecker()

# Run all checks
results = checker.run_all_checks(carousel)

# Generate report
report = checker.generate_report(results)
print(report)

"Content OS: Statins myth-busting - include carousel"

"Turn my latest newsletter into Instagram carousel slides"

output/carousels/GLP-1-for-weight-loss/
├── slide-01.png  # Hook
├── slide-02.png  # Tips
├── slide-03.png  # Tips
├── slide-04.png  # Tips
├── slide-05.png  # Tips
├── slide-06.png  # Tips
├── slide-07.png  # Stats
├── slide-08.png  # CTA
├── caption.txt   # Suggested captions
├── alt-text.txt  # Accessibility descriptions
├── hashtags.txt  # Recommended hashtags
└── report.txt    # Quality check report

pillow>=10.0.0      # Image rendering
pydantic>=2.0.0     # Data validation
plotly>=5.0.0       # Charts (optional)
kaleido>=0.2.0      # Plotly export (optional)