Evidence Dev

Evidence takes a code-first approach to BI — write SQL queries and Markdown, and Evidence builds interactive charts and tables into a deployable web app. Reports are version-controlled in git, deployable to Vercel/Netlify/any static host, and update on a schedule from database sources. For agent-assisted data workflows: generate Evidence reports programmatically or connect agents to Evidence's query outputs. No API key needed for the framework itself — it compiles to a web app. Confidence is docs-derived.

No REST API — Evidence is a static site generator. Data flow: SQL queries in .sql files → component bindings in .md files → Evidence builds the report UI. Connected sources: BigQuery, Snowflake, Postgres, MySQL, DuckDB, SQLite, Parquet files. Build: npm run dev for local preview, npm run build for static output. Deploy: Vercel, Netlify, GitHub Pages, or Evidence Cloud. Agent integration: can generate or modify .sql/.md files programmatically; build process produces static HTML.

No Evidence API key — open-source framework. Database credentials configured per source (environment variables or source-specific config files). Evidence Cloud: account auth for hosted deployment. Open-source MIT license.

Build failures: compile-time SQL errors and schema mismatches caught during evidence build. Static output means no runtime failures post-deploy. Source database connectivity issues surface at build time. Evidence Cloud uptime tracked separately. Git-based deployment ensures reproducibility.

docs.evidence.dev covers getting started, data sources, components, deployment, and custom theming. Getting started: npm install evidence, npx evidence create, first report in under 15 minutes. Open-source MIT (GitHub: evidence-dev/evidence). Community via Evidence Discord. Good for engineering-forward data teams that prefer code over drag-and-drop.