v0.2.0
Released 2026-10-13. Still wraps kuva 0.5.0 — this release is about the SQL surface and the SDKs, not about new chart types.
GitHub Release · Commits since v0.1.0
What changed
A second argument for every function. Each render function now also takes an optional JSON object
mapping dataset names to arrays of rows, so the chart config and the data can be built by different
things — the config by a host SDK, the data by SQL. A series names a dataset with dataset and maps its
columns with encode; the one-argument form is unchanged.
SELECT kuva_render(
'{"series":[{"type":"scatter","dataset":"sales","encode":{"x":"revenue","y":"profit"}}]}',
to_json({'sales': array_agg({'revenue': revenue, 'profit': profit})})
) FROM sales;
JSON Schema, generated from the Rust types. The schema under sdk/json_schema/ is rendered from the
same structs that parse the spec, so it cannot drift from the implementation.
SDKs. TypeScript types are generated from that schema (sdk/ts/), and Java (sdk/java/) and Kotlin
Multiplatform (sdk/kotlin/, including Android) models are generated from an OpenAPI 3.1 wrapper of it.
Each SDK has an integration test that loads a real release build and renders through it.
Docs. The gallery was regrouped by category, and the development notes moved onto the documentation site.
Breaking changes
- Rows are always objects. The
[x, y]positional form is gone: adataentry must name its fields ({"x": 1, "y": 2}). - Custom values are objects too. Inline
palettecolours,labelsandbrickentries that used to be bare values are now objects, matching the named forms they sit next to.
Specs written for 0.1.0 that only used object rows and named options keep working.