Mobile AI Assistant & UI Optimization

The Synalux portal features a global AI assistant and mobile-first input optimizations designed for speed and offline-resilience during clinical encounters.

Global AI Chat Assistant

The AI Chat is anchored via a Floating Action Button (FAB) across the entire platform — available on every page.

Model Routing

Cloud tiers route through OpenRouter with automatic fallback:

• Free — Claude Haiku 3.5 (fast, cost-effective)
• Standard / Advanced — Claude Sonnet 4 (primary), Claude Haiku 3.5 (fallback)
• Enterprise — Claude Sonnet 4 (primary), Claude Opus 4 selectable
• Gemini models — Gemini 2.5 Flash, Gemini 2.5 Pro, Gemini 3 Flash/Pro Preview available via direct Google API on Advanced+ tiers
• Local — prism-coder:14b via Ollama (free, offline, all tiers)

Capabilities

• Clinical context — the assistant understands the current page context (patient chart, billing, scheduling) and tailors responses accordingly
• Research synthesis — queries PubMed, ERIC, and Semantic Scholar for evidence-based answers with inline citations
• Code assistance — code generation and debugging for developers using Prism Coder
• Conversation history — chat sessions retain history in DOM state, so context persists when minimizing the chat to check patient data
• Voice input — dictate messages using the device microphone with WASM Whisper transcription

Mobile-First Design

• Safe area insets — UI accounts for env(safe-area-inset-bottom) on iOS and Android, so inputs never overlap the home indicator
• Touch-optimized — tap targets meet 44px minimum for medical device usability guidelines
• Responsive layout — adapts from phone to tablet to desktop with appropriate information density

Native Predictive Text

All Synalux inputs use native OS-level predictive text instead of custom JavaScript DOM injections:

• Uses spellCheck and autoCorrect for OS-native suggestions
• HIPAA-compliant — keystroke data stays on the device hardware, never piped to third-party AI ghost-text servers
• Works with specialized medical dictionaries when configured in the device's OS settings

Offline AI

When no network is available, the local Ollama integration provides full AI chat capability:

• prism-coder:14b model runs entirely on-device
• No data leaves the device — complete HIPAA compliance
• Available on all tiers at no cost
• Requires Ollama installed locally (macOS, Linux, or Windows)

VS Code Extension

The same AI assistant is available as a VS Code Extension for developers who prefer to work in their editor.