TriageDesk — AI-guided patient intake & red-flag triage for busy clinics.
In crowded outpatient departments the first minutes go to queue guessing while cardiac chest pain, stroke signs or meningitis wait unrecognized in line. TriageDesk gives the front desk — or the patient — a 60-second structured intake that surfaces red flags FIRST.
- Bilingual by design: full English + বাংলা interface, one-tap switch
- Free-text symptom extraction with negation guard ("no cough" / "কাশি নেই" correctly ignored) over a 65-symptom taxonomy
- Deterministic red-flag engine: 33 transparent rules (FAST stroke signs, cardiac patterns, meningitis triad, anaphylaxis, infant fever, obstetric emergencies) — every fired rule shown in plain language, no black box
- Context-aware escalation: age bands, pregnancy, chronic conditions
- Printable one-page doctor summary for the clinic record
- Optional BYO LLM (local Ollama) enriches extraction; triage logic never depends on it
- Zero infrastructure: static files, runs offline, no data leaves the device
Verified: 110/110 stress-harness samples across 10 test angles, 26/26 regression cases, 0 console errors in full browser flows.
Live demo: https://rudrasarker.is-a.dev/TriageDesk/ Repo: https://github.com/rudra496/TriageDesk Informational triage aid — not a diagnosis.
Built With
- bengali
- css
- healthcare
- html
- javascript
- ollama
- triage
