Agent Skills: Sankey Diagram Creator

Create interactive Sankey diagrams for flow visualization from CSV, DataFrame, or dict data. Supports node/link styling and HTML/PNG/SVG export.

UncategorizedID: dkyazzentwatwa/chatgpt-skills/sankey-diagram-creator

Skill Files

Browse the full folder contents for sankey-diagram-creator.

Download Skill

Loading file tree…

sankey-diagram-creator/SKILL.md

Skill Metadata

Name
sankey-diagram-creator
Description
Create interactive Sankey diagrams for flow visualization from CSV, DataFrame, or dict data. Supports node/link styling and HTML/PNG/SVG export.

Sankey Diagram Creator

Create interactive Sankey diagrams to visualize flows, transfers, and relationships between categories. Perfect for budget flows, energy transfers, user journeys, and data pipelines.

Quick Start

from scripts.sankey_creator import SankeyCreator

# From dictionary
sankey = SankeyCreator()
sankey.from_dict({
    'source': ['A', 'A', 'B', 'B'],
    'target': ['C', 'D', 'C', 'D'],
    'value': [10, 20, 15, 25]
})
sankey.generate().save_html("flow.html")

# From CSV
sankey = SankeyCreator()
sankey.from_csv("flows.csv", source="from", target="to", value="amount")
sankey.title("Budget Flow").generate().save_html("budget.html")

Features

  • Multiple Input Sources: Dict, DataFrame, or CSV files
  • Interactive Output: Hover tooltips with flow details
  • Node Customization: Colors, labels, positions
  • Link Styling: Colors, opacity, labels
  • Export Formats: HTML (interactive), PNG, SVG
  • Auto-Coloring: Automatic color assignment or custom schemes

API Reference

Initialization

sankey = SankeyCreator()

Data Input Methods

# From dictionary
sankey.from_dict({
    'source': ['A', 'A', 'B'],
    'target': ['X', 'Y', 'X'],
    'value': [10, 20, 15]
})

# From CSV file
sankey.from_csv(
    filepath="data.csv",
    source="source_col",
    target="target_col",
    value="value_col",
    label=None  # Optional: column for link labels
)

# From pandas DataFrame
import pandas as pd
df = pd.DataFrame({
    'from': ['A', 'B'],
    'to': ['C', 'C'],
    'amount': [100, 200]
})
sankey.from_dataframe(df, source="from", target="to", value="amount")

# Add individual flows
sankey.add_flow("Source", "Target", 50)
sankey.add_flow("Source", "Other", 30, label="30 units")

Styling Methods

All methods return self for chaining.

# Node colors
sankey.node_colors({
    'Income': '#2ecc71',
    'Expenses': '#e74c3c',
    'Savings': '#3498db'
})

# Use colormap for automatic colors
sankey.node_colors(colormap='Pastel1')

# Link colors (by source, target, or custom)
sankey.link_colors(by='source')  # Color by source node
sankey.link_colors(by='target')  # Color by target node
sankey.link_colors(opacity=0.6)  # Set link opacity

# Custom link colors
sankey.link_colors(colors={
    ('Income', 'Expenses'): '#ff6b6b',
    ('Income', 'Savings'): '#4ecdc4'
})

# Title
sankey.title("My Flow Diagram")
sankey.title("Budget Overview", font_size=20)

# Layout
sankey.layout(orientation='h')  # Horizontal (default)
sankey.layout(orientation='v')  # Vertical
sankey.layout(pad=20)  # Node padding

Generation and Export

# Generate the diagram
sankey.generate()

# Save as interactive HTML
sankey.save_html("diagram.html")
sankey.save_html("diagram.html", auto_open=True)  # Open in browser

# Save as static image
sankey.save_image("diagram.png")
sankey.save_image("diagram.svg", format='svg')
sankey.save_image("diagram.png", width=1200, height=800)

# Get Plotly figure object for customization
fig = sankey.get_figure()
fig.update_layout(...)

# Show in notebook/browser
sankey.show()

Data Format

CSV Format

source,target,value,label
Income,Housing,1500,Rent
Income,Food,600,Groceries
Income,Transport,400,Car
Income,Savings,500,Emergency Fund
Savings,Investments,300,Stocks

Dictionary Format

data = {
    'source': ['Income', 'Income', 'Income', 'Expenses'],
    'target': ['Rent', 'Food', 'Savings', 'Tax'],
    'value': [1500, 600, 500, 400],
    'label': ['Housing', 'Groceries', 'Emergency', 'Federal']  # Optional
}

DataFrame Format

import pandas as pd
df = pd.DataFrame({
    'from_node': ['A', 'A', 'B', 'C'],
    'to_node': ['B', 'C', 'D', 'D'],
    'flow_value': [100, 200, 150, 180]
})

Color Schemes

Available Colormaps

| Name | Description | |------|-------------| | Pastel1 | Soft pastel colors | | Pastel2 | Muted pastels | | Set1 | Bold distinct colors | | Set2 | Muted distinct colors | | Set3 | Light distinct colors | | Paired | Paired color scheme | | viridis | Blue-green-yellow | | plasma | Purple-orange-yellow |

Custom Colors

# Hex colors
sankey.node_colors({
    'Category A': '#1abc9c',
    'Category B': '#3498db',
    'Category C': '#9b59b6'
})

# Named colors
sankey.node_colors({
    'Revenue': 'green',
    'Costs': 'red',
    'Profit': 'blue'
})

CLI Usage

# Basic usage
python sankey_creator.py --input flows.csv \
    --source from --target to --value amount \
    --output flow.html

# With title and colors
python sankey_creator.py --input budget.csv \
    --source category --target subcategory --value amount \
    --title "Budget Breakdown" \
    --colormap Pastel1 \
    --output budget.html

# PNG output
python sankey_creator.py --input data.csv \
    --source src --target dst --value val \
    --output diagram.png \
    --width 1200 --height 800

CLI Arguments

| Argument | Description | Default | |----------|-------------|---------| | --input | Input CSV file | Required | | --source | Source column name | Required | | --target | Target column name | Required | | --value | Value column name | Required | | --output | Output file path | sankey.html | | --title | Diagram title | None | | --colormap | Color scheme | Pastel1 | | --link-opacity | Link opacity (0-1) | 0.5 | | --orientation | 'h' or 'v' | h | | --width | Image width (PNG/SVG) | 1000 | | --height | Image height (PNG/SVG) | 600 |

Examples

Budget Flow

sankey = SankeyCreator()
sankey.from_dict({
    'source': ['Salary', 'Salary', 'Salary', 'Salary', 'Savings', 'Investments'],
    'target': ['Housing', 'Food', 'Transport', 'Savings', 'Investments', 'Stocks'],
    'value': [2000, 800, 500, 1000, 700, 700]
})
sankey.title("Monthly Budget Flow")
sankey.node_colors({
    'Salary': '#27ae60',
    'Housing': '#e74c3c',
    'Food': '#f39c12',
    'Transport': '#3498db',
    'Savings': '#9b59b6',
    'Investments': '#1abc9c',
    'Stocks': '#34495e'
})
sankey.generate().save_html("budget_flow.html")

Energy Flow

sankey = SankeyCreator()
sankey.add_flow("Coal", "Electricity", 100)
sankey.add_flow("Gas", "Electricity", 80)
sankey.add_flow("Solar", "Electricity", 30)
sankey.add_flow("Electricity", "Industry", 120)
sankey.add_flow("Electricity", "Residential", 60)
sankey.add_flow("Electricity", "Commercial", 30)

sankey.title("Energy Flow")
sankey.node_colors(colormap='Set2')
sankey.link_colors(by='source', opacity=0.4)
sankey.generate().save_html("energy.html")

User Journey

sankey = SankeyCreator()
sankey.from_dict({
    'source': ['Landing', 'Landing', 'Product', 'Product', 'Cart', 'Cart'],
    'target': ['Product', 'Exit', 'Cart', 'Exit', 'Checkout', 'Exit'],
    'value': [1000, 200, 600, 200, 400, 100]
})
sankey.title("User Journey")
sankey.node_colors({
    'Landing': '#3498db',
    'Product': '#2ecc71',
    'Cart': '#f1c40f',
    'Checkout': '#27ae60',
    'Exit': '#e74c3c'
})
sankey.generate().save_html("journey.html")

Multi-Level Flow

# Three-level hierarchy
data = {
    'source': [
        # Level 1 -> Level 2
        'Revenue', 'Revenue', 'Revenue',
        # Level 2 -> Level 3
        'Products', 'Products', 'Services', 'Services'
    ],
    'target': [
        'Products', 'Services', 'Other',
        'Electronics', 'Software', 'Consulting', 'Support'
    ],
    'value': [500, 300, 50, 300, 200, 200, 100]
}

sankey = SankeyCreator()
sankey.from_dict(data)
sankey.title("Revenue Breakdown")
sankey.generate().save_html("revenue.html")

Dependencies

plotly>=5.15.0
pandas>=2.0.0
kaleido>=0.2.0

Limitations

  • Static image export (PNG/SVG) requires kaleido package
  • Very complex diagrams with many nodes may be hard to read
  • Node positions are auto-calculated (limited manual control)
  • Link colors can only be uniform or by source/target