System Architecture
Multi-layer AI-native distributed system. Multi-agent orchestration with lifelong memory and safety-first design.
Client Layer
Next.js Web (Dashboards)React Native / Expo (Child App)
API Gateway
Auth (Clerk/Supabase)RBACRate LimitingAudit Logging
AI Orchestration
Intent ClassificationAgent RoutingResponse ComposerMulti-Model Routing
Agent Layer
Learning AgentEmotion AgentRoutine AgentSocial AgentExecutive Function AgentSpeech Agent
Memory Engine
PostgreSQL (structured)pgvector (semantic)Emotional taggingLong-term memory
Safety Layer
Content moderationSafeguarding detectionRisk escalationParent alerts
35+ AI Agents
Learning Agent
Emotion Agent
Routine Agent
Social Agent
Executive Function Agent
Speech Agent
Sleep Agent
Safety Agent
Memory Agent
Guardian Agent
Homework Agent
Nutrition Agent
Exercise Agent
Travel Coach
Career Coach
Teen Coach
Habit Coach
Reflection Coach
Medication Agent
Confidence Coach
Friendship Coach
Emergency Coach
Transition Coach
Shopping Coach
Cooking Coach
Cleaning Coach
University Coach
Interview Coach
Relationship Coach
Adult Coach
Sensory Agent
Behaviour Agent
Communication Agent
Parent Agent
Teacher Agent
Therapist Agent
Technology Stack
Next.js + VercelExpo React NativeSupabase PostgreSQLpgvectorOpenAI / Anthropic / GeminiElevenLabs VoiceDeepgram STTStripePostHogSentryFirebase FCM
36-Month Roadmap
Phase 0
0–3 mo
Foundation, auth, base AI, design system
Phase 1
3–9 mo
MVP: learning, routines, emotions, parent dash
Phase 2
9–18 mo
Multi-agent, teacher/therapist dashboards
Phase 3
18–36 mo
Lifelong memory, institutional scale