Improve asset pipeline quality and AI icon consistency

[jadru]

Summary

• improve AI icon generation reproducibility and style consistency:
  • add seed and quality tier (fast / quality) support
  • preserve and return actual seed values
  • wire brand profile -> prompt config mapping with composeIconPrompt
  • add negative prompt support and "Similar" regenerate flow in UI
• fix asset generation correctness issues:
  • generate and package PWA icons (icon-192.png, icon-512.png)
  • update manifest icon paths and include icons under web/pwa
  • fix text icon contrast color usage for PNG rendering
  • improve SVG text-icon legibility and favicon base render quality
• improve output quality and compatibility:
  • WCAG-based contrast color algorithm
  • Android adaptive icon safe-zone handling
  • expanded splash screen target sizes
  • OG image text overflow handling with dynamic typography
  • PNG compression tuning per asset category
• harden generated snippets and pipeline reliability:
  • add escaping for TS/XML/HTML snippets
  • add image validation helper for critical generated assets
  • add structured logger usage in asset pipeline/storage routes
• add regression tests for colors, image utils, favicons, and code snippet escaping
• include missing public/fonts/Inter-Bold.ttf

Validation

• yarn exec tsc --noEmit
• yarn test

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
brand-kit	Ready	Preview, Comment	Feb 26, 2026 5:03pm

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c71eec7b71

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve Error details in structured logger payloads

writeLog serializes data with JSON.stringify as-is, but most call sites pass Error objects (for example in API and webhook catch blocks), and JSON.stringify(new Error(...)) drops message/stack to {}. This means production failure logs lose the actual exception details that operators need to diagnose incidents, which is a regression from the previous logger behavior that printed full error objects.

Useful? React with 👍 / 👎.