feat: add Kokoro neural TTS via /tts endpoint

fix: don't fall back to English voice for non-English locales

fix: force JSON output on quick-scan and remove redundant colour suffix

fix: remove raw category tag from scan long description

fix: scan/outfit description issues and richer shopping responses

feat: richer scan descriptions, outfit mode picker, shopping auto-pause, context-chat fix

fix: scan screen short mode shows full short_description, not first sentence

fix: wardrobe desc truncation, scan desc toggle, shopping camera blackout

fix: resolve 8 bugs found in full codebase audit

fix: shopping mode (#7)

fix camera issue

fix: groq integration & shopping mode

Merge branch 'claude/fervent-almeida-54401e' into main

feat(shopping): migrate to Groq/Llama-4-Scout + fix camera feed UX

Accessibility overhaul, Gemini-first architecture, multilingual support & bug fixes (#6)

feat(i18n): add Gujarati, Kannada, Malayalam, Punjabi translations

fix: mirror mode visual feedbackk

fix: preview mode

fix: language selection

fix: iOS camera issues

fix: camera stream issue

fix: shopping mode real-time activation

feat(i18n): add voice-first multilingual support with Indian languages

Merge branch 'origin/ayush' into main (prefer ayush changes)

Store rich wardrobe descriptions purpose-built for garment identification

Add garment identification feature to match physical items to wardrobe

Add ML-based frame detection using COCO-SSD for accurate preview mode

Add per-screen context chatbot with voice and text input

Add free-form LLM voice assistant with thinking state feedback

Remove algorithmic color engine; switch to pure LLM analysis

fixes

fix: nitin (#4)

Replace image storage with rich text descriptions; enforce single-item scanning

Add custom wardrobe naming; fix outfit recommendations to be fun and exact

Remove harmony/seasonal analysis; fix occasion and style to multi-label

Rebuild shopping mode with wardrobe-aware live feedback and follow-up questions

fixed camera bug in shopping mode

fix: improve camera guidance

fix: voice responses for audible mirror

fix: two pieces false positive and add more guidance

Clarify Scan vs Mirror distinction; add duplicate detection in wardrobe save

Fix OutfitScreen TDZ crash: move voice command useEffect after useCallback defs

Save all detected garments as separate wardrobe items instead of only the first

Remove inaccurate occasion badge from result screen

Add upload from gallery option to ScanScreen

Add visual guidance arrows/green zone; fix color name consistency in LLM prompt

Fix: move voice command useEffects after useCallback defs to prevent TDZ in prod

Fix production build crashes: LiveRegions module-level component, stale deps, remove dead code

Fix: LiveRegions initialization error in production build

Production build: remove guest mode, remove Anthropic, route outfit suggestions via Gemini backend