Lumen Dashboards

Lumen Dashboards Skill

Overview

Lumen is a declarative framework for creating data dashboards through YAML specifications. Build interactive data exploration dashboards without writing code.

Key Features

• No-code dashboards: Define everything in YAML
• Data pipelines: Sources → Transforms → Views
• Interactive exploration: Built-in filters and cross-filtering
• Component library: Reusable sources, transforms, views
• Live updates: Auto-reload and real-time data

When to Use Lumen

Use Lumen when: Building standard dashboards, working with non-programmers, rapid prototyping.

Use Panel when: Need fine-grained control, custom logic, novel interactions.

Quick Start

Installation

pip install lumen

Your First Dashboard

File: dashboard.yaml

sources:
  data:
    type: file
    tables:
      penguins: https://datasets.holoviz.org/penguins/v1/penguins.csv

pipelines:
  main:
    source: data
    table: penguins
    filters:
      - type: widget
        field: species

layouts:
  - title: Penguin Explorer
    views:
      - type: hvplot
        pipeline: main
        kind: scatter
        x: bill_length_mm
        y: bill_depth_mm
        by: species

Launch:

lumen serve dashboard.yaml --show

Core Concepts

1. Sources

Data sources provide tables for your dashboard.

Supported sources: File (CSV, Parquet, Excel, JSON), Database (PostgreSQL, DuckDB, SQLite), REST API, Intake catalogs.

sources:
  mydata:
    type: file
    tables:
      sales: ./data/sales.csv

See: Data Sources Reference

2. Pipelines

Pipelines define data flows: Source → Filters → Transforms → Views

pipelines:
  sales_pipeline:
    source: mydata
    table: sales
    filters:
      - type: widget
        field: region
    transforms:
      - type: aggregate
        by: ['category']
        aggregate:
          total_sales: {revenue: sum}

3. Filters

Add interactive controls:

filters:
  - type: widget
    field: category          # Dropdown select

  - type: widget
    field: region
    multiple: true           # Multi-select

  - type: widget
    field: date
    widget: date_range_slider  # Date range

4. Transforms

Process data in pipelines:

transforms:
  - type: columns
    columns: ['date', 'region', 'revenue']

  - type: query
    query: "revenue > 1000"

  - type: aggregate
    by: ['region']
    aggregate:
      total: {revenue: sum}

See: Data Transforms Reference

5. Views

Visualize data with various chart types:

views:
  - type: hvplot
    pipeline: main
    kind: line
    x: date
    y: revenue
    by: category

  - type: indicator
    pipeline: main
    field: total_revenue
    format: '${value:,.0f}'

  - type: table
    pipeline: main
    page_size: 20

See: Views Reference

6. Layouts

Arrange views on the page:

layouts:
  - title: Overview
    layout: [[0, 1, 2], [3], [4, 5]]  # Grid positions
    views:
      - type: indicator
        # View configs...

See: Layouts Reference

Common Patterns

Pattern 1: KPI Dashboard

pipelines:
  kpis:
    source: metrics
    table: data
    transforms:
      - type: aggregate
        aggregate:
          total_revenue: {revenue: sum}
          total_orders: {orders: sum}

layouts:
  - title: KPIs
    layout: [[0, 1, 2]]
    views:
      - type: indicator
        pipeline: kpis
        field: total_revenue
        format: '${value:,.0f}'
      - type: indicator
        pipeline: kpis
        field: total_orders
        format: '{value:,.0f}'

Pattern 2: Filtered Exploration

pipelines:
  explorer:
    source: mydata
    table: sales
    filters:
      - type: widget
        field: region
      - type: widget
        field: category
        multiple: true
    views:
      - type: hvplot
        kind: scatter
        x: price
        y: quantity
      - type: table
        page_size: 20

Pattern 3: Cross-Filtering

layouts:
  - title: Analysis
    views:
      - type: hvplot
        pipeline: main
        kind: bar
        x: category
        y: revenue
        selection_group: category_filter

      - type: hvplot
        pipeline: main
        kind: scatter
        x: price
        y: quantity
        selection_group: category_filter

Pattern 4: SQL Transform

transforms:
  - type: sql
    query: |
      SELECT region, category,
        SUM(revenue) as total_revenue,
        COUNT(*) as order_count
      FROM table
      WHERE date >= '2024-01-01'
      GROUP BY region, category
      ORDER BY total_revenue DESC

See: Examples for complete dashboard examples.

Python API

While Lumen is designed for YAML, you can also use Python:

from lumen.sources import FileSource
from lumen.pipeline import Pipeline
from lumen.views import hvPlotView
from lumen.dashboard import Dashboard

source = FileSource(tables={'sales': './data/sales.csv'})
pipeline = Pipeline(source=source, table='sales')
view = hvPlotView(pipeline=pipeline, kind='scatter', x='price', y='quantity')
dashboard = Dashboard(pipelines={'main': pipeline}, layouts=[view])
dashboard.servable()

See: Python API Reference

Configuration

Global Config

config:
  title: My Dashboard
  theme: dark  # or 'default', 'material'
  sizing_mode: stretch_width
  logo: ./logo.png

Themes

config:
  theme: material
  theme_json:
    palette:
      primary: '#00aa41'

Authentication

lumen serve dashboard.yaml \
  --oauth-provider=generic \
  --oauth-key=${OAUTH_KEY} \
  --oauth-secret=${OAUTH_SECRET}

Deployment

Development

lumen serve dashboard.yaml --autoreload --show

Production

panel serve dashboard.yaml \
  --port 80 \
  --num-procs 4 \
  --allow-websocket-origin=analytics.company.com

Docker

FROM python:3.11-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install -r requirements.txt
COPY dashboard.yaml data/ ./
CMD ["lumen", "serve", "dashboard.yaml", "--port", "5006", "--address", "0.0.0.0"]

See: Deployment Guide

Best Practices

1. Source Organization

# Use descriptive names
sources:
  sales_database:
    type: postgres
    tables: [orders, customers]

  inventory_files:
    type: file
    tables:
      stock: ./inventory.csv

2. Pipeline Reusability

pipelines:
  base_sales:
    source: data
    table: sales
    filters:
      - type: widget
        field: region

  summary_sales:
    pipeline: base_sales  # Extends base_sales
    transforms:
      - type: aggregate
        by: ['category']

3. Performance

# Limit data size for large tables
sources:
  bigdata:
    type: postgres
    tables:
      events: "SELECT * FROM events WHERE date >= '2024-01-01' LIMIT 100000"

4. User Experience

# Provide clear labels
filters:
  - type: widget
    field: region
    label: "Sales Region"

views:
  - type: indicator
    field: revenue
    title: "Total Revenue"
    format: '${value:,.0f}'

Troubleshooting

Dashboard Won't Load

# Check YAML syntax
python -c "import yaml; yaml.safe_load(open('dashboard.yaml'))"

# Run with debug logging
lumen serve dashboard.yaml --log-level=debug

Data Not Showing

• Verify data source path/connection
• Check table names match YAML config
• Ensure columns referenced exist in data

See: Troubleshooting Guide

Resources

Reference Documentation

• Data Sources Reference - All source types
• Data Transforms Reference - Complete transform reference
• Views Reference - All visualization types
• Layouts Reference - Layout patterns
• Python API Reference - Programmatic creation
• Examples - Complete dashboard examples
• Deployment Guide - Production deployment
• Troubleshooting Guide - Common issues

External Links

• Lumen Documentation
• Lumen Gallery
• GitHub Repository
• Community Discourse

Summary

Lumen enables rapid dashboard development through declarative YAML specifications.

Strengths: No Python code required, fast development cycle, reproducible specifications, built-in interactivity.

Ideal for: Fixed dashboard layouts, standard data patterns, non-programmer dashboard creators, rapid prototyping.

Consider alternatives when:

• Need custom application logic → Panel Dashboards
• Need ad-hoc exploration → Lumen AI
• Building novel interactions → Panel Dashboards

Related Skills

• Lumen AI - Conversational data exploration
• Panel Dashboards - Custom Python dashboards
• Plotting Fundamentals - Quick plotting with hvPlot