PhraseGarden 0.1.0-preview.4

PhraseGarden 0.1.0-preview.4 is a public prerelease of the static, local Prompt Studio. See the release notes and qualification record for scope, limitations, and exact local evidence.

PhraseGarden 0.1.0-preview.3

PhraseGarden 0.1.0-preview.3 is a public prerelease of the static, local Prompt Studio. See the release notes and qualification record for scope, limitations, and exact local evidence.

PhraseGarden 0.1.0-preview.2 — usability update

PhraseGarden 0.1.0-preview.2 is a plain-language and zero-friction usability update to the deterministic public Preview.

Use PhraseGarden: https://jkolantree.github.io/phrasegarden/

What changed

• Create a prompt directly from Home with safe defaults; optional settings remain available.
• Plain-language choices explain what the prompt will do and where to use it.
• Primary actions stay visible in initial desktop and narrow mobile viewports.
• Copy and Download report visible outcomes.
• Edited prompts survive app navigation and require confirmation before replacement.
• Destination handoff and privacy guidance now distinguish Written and Voice use.

Deterministic boundary

The compiler, profiles, recipes, pair pack, generated instructions, canonical prompt bytes, and compiler provenance are unchanged at 0.1.0-preview.1. This release advances the application/package identity to 0.1.0-preview.2.

Verification

• 271/271 unit and snapshot tests
• Both TypeScript configurations, Vite build, release audit, and deterministic-domain scan
• 9/9 sequential Playwright/axe journeys on Microsoft Edge and Ubuntu Chromium
• Corrected GitHub Actions run: https://github.com/jkolantree/phrasegarden/actions/runs/30097157552
• The first Ubuntu run failed closed on a below-fold desktop action; the captured assertion remained unchanged and the corrected run passed.
• 9/9 checksum targets and 3/3 manifest-bound archive files
• Production Pages HTML, CSS, and JavaScript match the qualified manifest exactly
• Production English-to-Japanese Written creation, copied text, and downloaded text passed

Qualified artifact

phrasegarden-0.1.0-preview.2-pages.zip — 44,006 bytes
SHA-256: F76C665A2919B5403627667CB76D72015B0A4E79C0D95D653B4AD9EB6D4C15CF

Manifest SHA-256: 54EA72417BFCC34246E937A427DD0C7BFD17132FA9EB8C42D7B173D51CEBAE38

The attached manifest records every Pages file. SHA256SUMS qualifies both release archives, both manifests, and the five canonical prompt samples.

Preview limitations

External English-Japanese linguistic review remains incomplete. Generic output intentionally contains no pair-specific guidance. The interface is English-only. Interpreter, persistence, sharing, offline refresh, and a completed manual assistive-technology matrix remain deferred.

Code is MIT-licensed. Maintained prompt and content material is CC BY 4.0-licensed.

PhraseGarden 0.1.0-preview.1

PhraseGarden 0.1.0-preview.1 is the first public Preview of the deterministic, local-first Prompt Studio.

Use PhraseGarden: https://jkolantree.github.io/phrasegarden/

Included

• English→Japanese and Japanese→English Preview guidance
• Conservative Generic generation for every other bundled direction
• Written Translator and Live Voice Coach
• Readable behavior summaries and generated prompts
• Local copy, explicit editing, regeneration, and UTF-8 text download
• No backend, account, telemetry, runtime AI, or source-text collection

Verification

• 270/270 unit and snapshot tests
• Both TypeScript configurations and production build
• 4/4 sequential Playwright/axe browser journeys
• 7/7 checksum targets and 3/3 manifest-bound archive files
• Production Pages HTML, CSS, and JavaScript match the qualified local byte lengths and SHA-256 values

Qualified artifact

phrasegarden-0.1.0-preview.1-pages.zip — 40,963 bytes
SHA-256: A9952F91DF731D6C4D3785DBFB24513D6477B571F659E3A3C1EB5E65F5618357

The attached manifest records each Pages file. SHA256SUMS qualifies the archive, manifest, and five sample prompts.

Preview limitations

External English↔Japanese linguistic review is incomplete. Generic output intentionally contains no pair-specific linguistic guidance. The interface is English-only. There is no persistence, sharing, service worker, runtime AI, or completed manual assistive-technology matrix.

Code is MIT-licensed. Maintained prompt/content material is CC BY 4.0-licensed.