evidence

name	evidence
description	Build and deploy Evidence.dev data apps for static portfolio dashboards. Use when the selected visualization path is Evidence and the dataset is CSV, Excel, Parquet, or DuckDB. Deploys to Netlify/Vercel with no server required.
license	MIT
argument-hint	[portfolio-goal] [dataset-type]
metadata	{"author":"data-visualization-kit","version":"1.0.0"}

Evidence Skill

Production-ready Evidence.dev delivery for static portfolio dashboards, browser-side DuckDB analysis, markdown-first page authoring, and free-tier static deployment.

When to Use

• building a static portfolio dashboard from CSV, Excel, Parquet, Kaggle, or DuckDB-friendly files
• deploying to Netlify or Vercel with no always-on server
• delivering a portfolio piece that should run locally with npm run dev
• keeping SQL close to the dataset while authoring presentation in markdown pages

When Not to Use

• the project needs a live SQL database connection or heavier stakeholder BI workflows, better matched to Metabase
• the project is operational, alert-centric, or time-series first, better matched to Grafana
• the team is already committed to a legacy Superset estate that must be maintained

Evidence Mindset

The 10 Commandments of Evidence Delivery:

1. Data contract before dashboard polish
2. SQL in sources, presentation in pages
3. One source per logical dataset
4. DuckDB SQL is standard - keep queries portable
5. Components are markdown-first, not config-first
6. Filters use input components, not server-side params
7. Build before deploy - never push unbuilt
8. Keep credentials in env vars, never in source files
9. npm run dev parity with production - no surprises
10. If live DB is needed, route to Metabase

Reference Navigation

• references/evidence-core.md - project structure, sources, pages, components, DuckDB SQL
• references/evidence-deploy.md - local dev, build, deploy, env vars, static hosting
• references/evidence-portfolio-pattern.md - kit integration pattern for per-project Evidence workspaces

Quick Decision Matrix

Need	Choose
Static portfolio dashboard from CSV/Excel/Kaggle	Evidence
Live SQL BI dashboards for stakeholders	Metabase
Operational or time-series monitoring	Grafana
Legacy Superset estate maintenance	Apache Superset

Key Best Practices

Data and Sources:

• keep source SQL isolated per logical dataset
• prefer portable DuckDB SQL over one-off engine-specific tricks
• treat npm run sources as part of the data contract, not optional prep

Pages and Components:

• keep business narrative in pages and transformation logic in sources
• use built-in Evidence components before inventing custom chart wiring
• keep filters aligned to user questions, not raw column names

Delivery:

• verify npm run dev locally before any deployment handoff
• build with npm run sources && npm run build before publishing
• keep secrets in environment variables only

Resources

• Evidence.dev: https://evidence.dev/
• Evidence docs: https://docs.evidence.dev/