rmerk splits the medical timeline into clinical lenses

The fork is aimed at medical-malpractice work, where teams sift through hospital record exports that can run thousands of pages. rmerk's earlier chronology collapsed everything into a single timeline; this update slices the same extracted events into a medication administration record view and a vitals view - the angles clinicians and expert witnesses actually argue from. The default extraction engine also moves to Kimi K2.6, an NVIDIA-hosted vision model, tuned for the multi-hour runs these record sets demand.

A separate fix stops a memory leak that bled browser memory every time a reviewer zoomed, resized, or switched documents inside a 3,000-page Epic export. Memory now plateaus during a session, though rmerk is candid the deeper problem - every page rendering at once - still needs separate work.