Vizardry profiles any dataset, brainstorms 12 to 20 chart options across three creative bands, draws every one live with D3, and only ships the report when the verifier runs green.
>/datastorm innovation-graph/One slash command runs the whole pipeline. You bring a dataset; Vizardry brings the corpus, the taste, and the proof.
Reads your CSV, JSON, parquet, or SQLite. Types every column, finds the shape signature, and computes the aggregates the page will cite.
12 to 20 options across conventional, analytical, and creative bands. Every option names a real exemplar from the 212-chart corpus.
Every option is rendered from your actual data, not described. Failure modes on the cards are observed, never speculative.
Eleven hard checks run the page's own scripts in jsdom: charts drew, cards complete, exemplars exist. Green or it does not ship.
Translated from Observable notebooks and Bostock's algorithm gists into one React + D3 dialect. Retrieval picks the two or three closest before a single line is written.
rows drift in opposite directions · every tile is a looping capture from the gallery
The verifier loads the finished report in jsdom and runs its own scripts. Every chart must draw from your real data. Every option card must carry all nine fields. Every cited exemplar must exist in the corpus.
Eleven hard checks, each one mutation-tested: the suite proves the verifier fails a page broken in exactly that way, with exactly that message.
A real datastorm run over the GitHub Innovation Graph: six years of developers, repositories, languages, and cross-border collaboration for every economy on GitHub — 386,944 rows across 8 files, 16 drawn options, every number measured from the data. This is the actual shipped report, not a mockup.
open the live report → vizardry.codeblackwell.ai