Demo mode — all data is simulated. Switch to Live to call the real detection pipeline.
Resident
Mrs Tan, 74
Detection: Loading…
Personalised 7-day baseline · not population norms
Risk index
Loading
Risk score breakdown Total: — / 120.5 pts
Weight-based scoring · no ML classifier · each point is traceable to a specific event
CUSUM drift monitor Stable
Accumulates small deviations — catches gradual decline a z-score would miss
Today's events
Last 7 days
Detection engineFNR = 0.000
LangGraph · 5 specialist agents · hybrid RAG · weight-based scoring
3-agent benchmark · 20 deterministic scenarios
AgentF1FNRp95 latency
LangGraph1.0000.0005724ms
Custom1.0000.0003ms
LangChain1.0000.0002ms
Agent routing — current run
FallAgentskippedno fall detected
MedAgentfiredpill_taking timing +8pts
ChronicAgentco-firedMetformin history present
RoutineAgentfirednap +20min vs baseline
SummaryAgentalwaysmerges specialist results
Hybrid RAG · BM25 + ChromaDB · RRF k=60
47 clinical facts · MRR 0.960
Ask the agentsLLM explains what detection found
Why did RoutineAgent fire? Explain the CUSUM trend What caused the 8pt timing score?
Detection runs continuously.
Ask why something was flagged.
Agents running
Connecting…