examples/

README

Examples

Two layers, two jobs.

• minimal/ — learn by copying. Small, single-mechanism notebooks (15–65 lines each) meant to be copied into your own project. Start here if you are building your first real notebook.
• The corpus (this directory) — see what it can do. Larger notebooks that each put one design property under pressure — direct manipulation, linked views, typed units, document rendering, live feeds, native fan-out. 38 of them run live in the browser at go-notebook.dev.

New here? Read these five first, in order:

#	Example	What it teaches	Run
1	minimal/hello	the whole model on one screen: input → derived, wired by name+type	go tool notebook run ./examples/minimal/hello
2	tempconv	a complete notebook end-to-end: controls, an SVG view, a test	go tool notebook run ./examples/tempconv
3	minimal/htmlcard	rich output the easy way (text/html), and the optional nb package	go tool notebook run ./examples/minimal/htmlcard
4	minimal/headless	the differentiator: the same file as a browser app and a --headless --json batch job	go tool notebook run ./examples/minimal/headless
5	capacity	a real interactive planning tool that is also a schedulable job	go tool notebook run ./examples/capacity

Then browse minimal/ for a copy-paste recipe per mechanism, or the corpus below for fuller demonstrations.

The corpus by theme

Every corpus notebook is an ordinary Go package — read it as Go. Those marked live run in the browser (their call graph touches no net/os/cgo, so they compile to WASM); the rest are the same file built as a native binary.

Controls and interaction — curvefit (draggable points), anscombe, boundary, fourier, kmeans.

Linked views — lotka, spectrogram, nbody (an invariant exposes "running is not passing").

Document output (HTML) — invoice, simpson (a table is the right medium for Simpson's paradox), punchcard.

Systems and HPC — critpath, backfill, roofline, amdahl, fattree, cachepolicy, latencybw, aimd, bdp, reliability, retrystorm, fleet.

Advanced browser integration — surface (a raw HTML/JS escape hatch), gpulife (WebGPU for browser-side parallelism), turing (native fan-out vs. single-threaded browser execution).

Live feeds — sensorfeed, homefeed, tickerfeed, apifeed (each is a pure notebook plus a driver/ that pushes data through the set port; see docs/live-feeds.md).

Native or heavyweight — lego, seam, portfolio, taxi (DuckDB via cgo), queue.

Numerical and statistical — bayes (reversible recomputation), clt, simpson, curvefit, mandelbrot, percolation, pid, little (dimensional correctness through Go types), consistenthash.

For the full argument the corpus makes — and the design findings that came out of building it — see the paper.