v0.1.0 — first public release

First public release of vizier — generate and critique journalistic data visualizations.

pip install datavizier          # distribution name; import + CLI are both `vizier`

The distribution is published to PyPI as datavizier (the bare vizier name was already taken); everything you import and run is still vizier — import vizier, vizier guide …, vizier mcp.

Highlights

• Computable toolkit (deterministic, no LLM, no keys): validate, suggest-palette, suggest-ramp, ink, analyze, recommend-form. Every suggestion is validated before it's returned — a faithful port of the dataviz color method (Machado-2009 CVD transforms, OKLCH, WCAG).
• Chart-pattern library: 43 chart forms across the nine FT Visual Vocabulary families, each with when-to-use / when-not / alternatives / common-mistakes / reading-checklist, queryable via BM25.
• Corpus-backed critique (optional [critique] extra): retrieval-augmented judgment of a chart against a corpus of critical writing, routed through somm.
• MCP server (vizier mcp): the toolkit + corpus query exposed to any MCP client.
• Self-contained chart-forms guide (docs/reader/) and a GitHub Pages site.

Light core + optional extras ([search], [critique], [ingest]); proprietary corpus prior art is pluggable via VIZIER_EXTENSION_DBS and never ships in the public package.

See CHANGELOG.md for the full list.