Skip to content

v0.2.7 — dependency maintenance + vitest 5

Choose a tag to compare

@jaypetez jaypetez released this 08 Sep 17:24
ec81cd9

A dependency-maintenance release rolling up the ten pending updates, including
the vitest 4 → 5 major. No runtime behaviour changed.

Changed

  • Runtime dependencies: hono 4.13.5 → 4.13.7 (api, web),
    @anthropic-ai/sdk 0.122.0 → 0.124.0 (core), @hono/zod-validator 0.9.0 →
    0.9.1 (api), and @clack/prompts 1.7.0 → 1.8.0 (cli).
  • Dev tooling: vitest and @vitest/coverage-v8 4.1.11 → 5.0.0,
    @types/node 26.2.0 → 26.5.0, @playwright/test 1.62.1 → 1.63.0,
    @changesets/cli 3.0.1 → 3.0.2, and lint-staged 17.4.1 → 17.5.0.
  • Vitest 5 needed no config migration — no vitest.config.* /
    vite.config.* changes and no test changes. Verified across the full CI
    matrix (Ubuntu/macOS/Windows × Node 22/24). (#94)

Fixed

  • stride profile stopped type-checking under @clack/prompts 1.8.0. That
    release pulls @clack/core 1.5.0, which retypes isCancel from
    (value) => value is symbol to (value) => value is typeof CANCEL_SYMBOL.
    The old signature narrowed away all symbols, collapsing a boolean | symbol
    union to boolean after the guard; the new one narrows only that unique
    symbol, and subtracting it from the broad symbol still leaves symbol. The
    PAR-Q screening loop therefore needed an explicit boolean assertion at one
    call site — sound, because isCancel guards with an early return immediately
    above and the cancel symbol is the only symbol confirm() can resolve to.
    Type-level only.

Still open

esbuild 0.27.7
(GHSA-g7r4-m6w7-qqqr, low,
dev scope) remains deferred, unchanged from 0.2.6: tsup@8.5.1 pins
esbuild: ^0.27.0, and the vulnerable path is esbuild's own development server,
which tsup never starts.

Full changelog: v0.2.6...v0.2.7