feat: add paper abstract, SVG architecture diagram, and full results tables

fix: replace all fake 99.9% metrics with honest 96.68% val accuracy

fix: animate progress bar during server inference + speed up CPU pipeline

fix: render Mamba & Layers tabs with real data from server.py

fix: invalidate old caches missing deep_analysis field

fix: scope XAI tab selector to #rp1 to not break Mamba/Layers tabs

feat: redesign Mamba & Layers tabs with real computed visualizations

fix: update URLs to retvim-demo2 space

chore: trigger fresh space rebuild 2026-03-28

fix: add app_port: 7860 to HF Space README frontmatter

chore: force fresh build from latest commit

fix: correct API_BASE URL (khamir->earther) + robust html path lookup

chore: trigger rebuild after force push

fix: add retvim-final.html + images/ (all binaries via LFS), fix numpy+images errors

fix: explicit pip installs in Dockerfile - no shell grep tricks

perf: use CPU-only PyTorch wheel to cut build time from 30min -> ~5min

fix: add COPY real_weights.pth to Dockerfile so model weights land in container

feat: push real model weights + updated source files to HF Space

fix: replace libgl1-mesa-glx with libgl1 for Debian Trixie compatibility + correct HF Space URL

fix: valid HF colorTo (indigo)

Initial commit: RetViM retinal disease classifier