ai

A chatbot needs both a quick summary and a full explanation from one LLM generation. The answer: marker-based splitting in the SSE stream, with separate cache storage and mid-stream parsing.

The source list and the chat fought over one overlay. Two panels with different z-index plus render-time visibility ended the fight.

Tag-matched related articles kept missing the point. Swap in cosine similarity, add a TLDR box, and make search live, all fail-safe against an old API.

Tired of the stiff left border on my blog TOC, I swapped it for a Tailwind gradient spine that fades downward. Pure CSS utility, plus a note on the bg-gradient-to-* rename in v4.

Orange looked great in dark mode but washed out in light. A split accent, emerald-600 in light and mint in dark, clears the WCAG contrast bar in both.

Flat gray tag chips all carry equal weight. A 10% primary tint plus a hairline inset ring makes them feel alive and clickable again.

Sidebar headings at 11px uppercase dim gray were the hardest text to read. Switching to 13px bold normal-case plus icons raised contrast and fixed asymmetry.

Swapping a chip border for a background nudged the whole row by 1px. The fix: an inset ring, because box-shadow never affects layout.

Complete guide to embedding YouTube, Mermaid, and Recharts in a Next.js blog without plugins. Boost engagement by 52% with consistent styling and dark mode.