We've given the entire app a visual refresh — every card, detail page, and table has been redesigned for consistency and clarity.

• Redesigned cards — Dataset, eval, and ground truth cards now share a clean portrait layout with icons, metadata rows, and separators. Dataset cards show which evals are attached. Eval cards show when they were last run.
• New runs table — Runs are now displayed in a proper sortable table with filters for eval, column, status, dataset, and date range. Click any row to see the full results.
• Cleaner detail pages — Eval and dataset detail pages use flat section headers instead of nested cards, giving everything more breathing room.
• Eval descriptions — Evals now have an optional description field so you can capture the intent behind each eval in plain language.
• Daily rate limit — To keep things fair, there's now a limit of 200 trace evaluations per day per user. You'll see a clear message if you hit it.