Update README.md

Update README.md

Card: add the LinkedIn launch post link

Card: both demo videos (Gradio app + hand-built UI) + GitHub link

Reviews refresh on load

Card + fresh screenshots: accounts (optional) and a public reviews wall

Fix account greeting alignment

SPA testimonials: real only, no seeds

Reviews form + cleanup (no seeds, no flags, no em dashes), aligned icons

Gradio: accounts + saved history, reviews wall, models showcase

README: state up front that the live Space is a working Gradio app (+ /classic alternate)

Card + SPA + diagram: text on Qwen, MiniCPM-V-2 (vision) is the OpenBMB MiniCPM build

Text on Qwen (Q&A + understanding); MiniCPM5-1B not reliable for structured output

Card + SPA + diagram: MiniCPM5-1B understands appliances, Qwen answers Q&A (two OpenBMB MiniCPM)

Q&A -> Qwen, appliance understanding -> MiniCPM5-1B (both genuinely used)

Text model -> MiniCPM5-1B; now two OpenBMB MiniCPM models (text + vision)

Gradio: warm models on load + disclaimer; footer links (classic UI, GitHub, SoroTTS)

Beautify the Gradio interface (atmosphere, hero, depth, no chrome)

Gradio app is now the Space interface (mounted at /); SPA kept at /classic

Space card: link the models and training datasets in metadata

Warm all models on open + loading disclaimer (scale-to-zero UX)

Space card: add demo video, expand tags, note scale-to-zero warm-up

Home models card -> MiniCPM-V-2; refreshed architecture diagram

Vision: MiniCPM-V-2 as primary (Best MiniCPM Build), Qwen2.5-VL kept as fallback

Logo color matches the hero heading (green-d) in both themes

Header on a rectangular card (logo, tagline, live, count)

Home page: models showcase (SoroTTS fine-tune) + GitHub and SoroTTS links

Update Space card: current models (SoroTTS, Qwen2.5-VL-3B), live screenshots, architecture, accurate prize mapping

Refresh assets: current live screenshots + new architecture/flow diagrams

Naturalize yo/ha/ig narration for the voice (numbers as words, units dropped); persistent /data audio cache

New-sizing reset + photo extraction callout; SoroTTS Yoruba fix + faster full-plan narration; diacritized yo/ha; 5-language i18n

Refresh article screenshots to the current app (home, 2D, 3D, written plan, dark mode)

Fix photo upload drop zone collapsing (label needed display:flex); center the prompt and thumbnails

Clean testimonials (clear dev/test entries + dedupe) and show each reviewer's language; pre-warm the voice

Pre-warm the voice on app entry so audio is ready sooner

Voice-first UX overhaul: instant written plan + loading state + clearer faster speech, first-run guided tour, asset cache-busting

Richer, more prominent animated background: rising sun with rotating rays, gradient-shaded panels/bulbs/leaves, twinkling sparkles, brighter in both themes

Smarter Q&A advisor (no brittle refusal, never blank, clean English for Yoruba/Hausa/Igbo), light/dark mode toggle, richer transitions and icons

Home page with sliding testimonials, full 5-language i18n across the whole UI, animated solar background, rating capture from all users, responsive pass

Full app: user accounts (signup/login), per-user sizing history, admin usage dashboard; local /data storage (durable via persistent storage), self-managed keys, no exposed credentials

Architecture diagram: FastAPI bespoke frontend (off-brand)

Off-brand: bespoke custom frontend (FastAPI + hand-built HTML/CSS/JS) replacing the Gradio surface; reuses the engine, generators, 3D scene and models

Add one-tap example scenarios (instant full result for first-timers and reviewers), responsive-verified

Add architecture + process-flow diagrams

Add architecture + process-flow diagrams

Refresh 3D screenshot (HUD + orbit)

3D upgrade: drag-to-orbit with front-arc clamp, sizing HUD chips, titled card, responsive canvas

Add app screenshots for the writeup

Fix: post-sizing explanation now follows the selected language (reliable templates for Pidgin/Yoruba/Hausa/Igbo)

Naija Solar: voice-first solar sizing in 5 languages, every model <=4B