finds.dev← search

// the find

evidence-dev/evidence

★ 6,429 · JavaScript · MIT · updated Feb 2026

Business intelligence as code: build fast, interactive data visualizations in SQL and markdown

Evidence turns markdown files with embedded SQL into static BI websites. You write `.md` files, drop SQL blocks in them, and it builds a Svelte app with charts and tables that runs entirely in the browser — DuckDB via WASM handles local query execution at runtime. It's aimed at data/analytics engineers who already live in SQL and git, and want dashboards they can version-control and deploy to Netlify like any other static site.

The DuckDB-WASM angle is the real differentiator: queries run client-side after the static build, so you get interactive filtering without a backend or API layer. Datasource breadth is solid out of the box — Postgres, Snowflake, BigQuery, Redshift, DuckDB, SQLite, CSV, Databricks, Trino, MSSQL all have connectors with tests. The git-native workflow (version control, PR reviews, CI/CD for dashboards) is a genuine improvement over click-and-pray Tableau/Looker. The VSCode extension with live preview makes the dev loop fast enough to actually use.

The README now leads with a tip to use 'Evidence Studio' (their commercial SaaS product) — a sign the OSS project may be sliding toward lead-gen rather than active development, and the last push being February 2026 while this is being evaluated in June 2026 supports that concern. Static builds are the wrong model for operational dashboards that need fresh data hourly or on-demand; you're either running expensive full rebuilds on a cron or accepting stale numbers. Large datasets get pre-aggregated at build time or shipped as big WASM-queryable blobs — there's no lazy server-side pagination. Templated pages and loop logic are squeezed into markdown syntax, which works fine until you need something the DSL doesn't cover, at which point you're fighting the abstraction.

View on GitHub → Homepage ↗

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →