[codex] Fix SwiftVoice privacy and speech robustness

[murdawkmedia]

Summary

Implements the SwiftVoice robustness and honest-privacy plan as an owner branch that supersedes PR #1 without carrying forward misleading local/offline claims.

What changed

• Refactored Web Speech API access behind typed, mockable helpers.
• Fixed transcript assembly so split recognition segments are joined with spaces.
• Added intentional-stop restart guards, inline unsupported/microphone error states, and timer start only after recognition starts.
• Replaced keystroke wording with voice-appropriate word/character stats and fixed mismatch accuracy tracking.
• Removed Tailwind CDN, Google Fonts, and import-map runtime CDN loading; added local Tailwind/PostCSS CSS pipeline.
• Rewrote README, metadata, and page copy around honest Web Speech privacy: no SwiftVoice backend/account/API keys/Murdawk collection, while browser speech behavior depends on the browser.
• Added unit tests, mocked Playwright e2e tests, npm scripts, and GitHub Actions CI.

Validation

• npm run typecheck
• npm test
• npm run build
• npm run test:e2e
• npm audit --json reports 0 vulnerabilities

Notes

The e2e tests mock speech recognition and do not access the real microphone. They also block third-party runtime requests to verify the app loads from same-origin assets only.

[cloudflare-workers-and-pages]

Deploying swiftvoice with    Cloudflare Pages

Latest commit:	29464fa
Status:	✅  Deploy successful!
Preview URL:	https://8562d5f0.swiftvoice.pages.dev
Branch Preview URL:	https://fix-swiftvoice-robustness-pr.swiftvoice.pages.dev

View logs