MOSTRAmi v0.8.0 Unreleased Source Candidate

New MOSTRAmi banner normalized for Play Store and site, new Play 512 icon, and Android raster/adaptive launcher for all densities.

Compact onboarding and new interactive guide with routes, feature catalog, QA checklist, and operations glossary of the installed build.

Instant shutter Freeze UX with artwork screening. Diagnostic HUD overlay dashboard (pipeline state, active engine, RAM, and connectivity) and safe scanning with 30s/35s timeout and native JNI deadlock resolution.

ML Kit pipeline for reading captions, parsing metadata, and passing context to the artwork analysis flow.

Contract-First architecture with InterpretationContract. The cloud text path uses one structured Gemini 3.6 generator, deterministic citation validation and a separate semantic/safety verifier. Android renders verified claim-level citations and each assistant message preserves its own contract task, grounding label and provenance.

Local download of Gemma 4 model variants is enabled and manageable directly in both Standard and Expert profiles. Local analyze/chat paths can now receive the rendered agent prompt containing the contract task and retrieved evidence.

Automatic vocal playback (Auto-Narration) of chatbot responses, with automatic cleaning of metadata tags and fluid interruption handling.

Native Clio Live Cloud targets the correct Firebase Google AI native-audio model, exposes continuous listening, voice response and barge-in, and is now available as a floating assistant across the main app surfaces. Live and TTS are mutually exclusive, local voice options stay hidden outside local/expert context, onboarding uses the Live path instead of legacy TTS, and chat releases active resources when the user returns home. Uses Firebase @PublicPreviewAPI, which could change before GA.

Kids (under 14, engaging tone with emojis) and Senior (high readability, strictly without emoticons) profiles covered by current product tests.

Italian and English UI coverage is stable. FR/DE/ES resources now include safe English fallbacks to keep lint clean while reviewed translations remain pending.

Photo saving disabled by default, Firebase diagnostics controlled by user consent, and AI disclosures clearly visible.

Local history, GDPR JSON export, scan deletion, and consent resets linked directly to user flows.

Multi-turn chat now includes an interactive artwork context panel grounded on the captured frame: pinch-to-zoom, tap a visible detail or detected point of interest, circle-select an area, and pass that crop into follow-up analysis. The June 28 pass gives more room to the artwork, moves reset and circle actions into compact icon controls, and renders Clio narration as readable rich text instead of raw markdown. AI disclosure, in-app reporting, and centralized free-tier limits remain active.

Italy-only catalog with Uffizi, Brera, Museo Egizio, Gallerie dell'Accademia, MANN, Capodimonte, and MAXXI. The screen now prioritizes official maps, optional distance, directions, and route choice without exposing distracting or unimplemented controls.

Each museum can expose curated experience packs with duration, audience, artworks, route steps, included assets, and a Clio opening line. Loading a pack creates an active visit, fills the wishlist, and sends the visitor directly to Camera to begin scanning. Packs are client templates in the current build; the signed production broker remains a launch gate.

Wear OS module with audio bridge service and remote controls. Real audio playback on smartwatch implemented but not yet validated on physical hardware.

The Scholar surface now reads as a premium dossier: guided process, source pack, critique, citation cards, note and share actions, and stronger hierarchy. Commercial positioning still needs billing and content QA before public sale.

The login button no longer fails silently, but real login requires a configured Web Client ID in Firebase.

PRO screen remains stable without offers, but Play Console products, RevenueCat offerings, and purchase QA are external.

Firebase Functions code passes syntax checks and preserves structured errors, but the production EU deployment, server-side App Check, rate limiting, and operational logging still need live verification.

Keystore uploads, CI secrets, public legal pages, final Data Safety, and internal Play Store track testing are required before production.

Real-time WebSocket synchronization of screens, centralized AI profile steering (Kids/Senior/Standard), and log-RSSI based BLE proximity warnings (>15m) with haptic vibration.

Art wishlist turns into riddles to solve by exploring rooms. Unlock works by scanning them, passing AI multiple-choice quizzes, and collecting digital stickers.

Complete post-visit redesign into a historical memories scrapbook. Vertical timeline with voice notes (.aac), customizable photos inserted via camera, XP statistics, and a Polaroid postcard generator.

Technical scaffolding for the cloud proxy workflow. Deep links and response simulations are implemented, but no real backend, no active Business APIs, and audio placeholders. Engineering has no fixed date and remains outside the closed-beta release path.