8. Symptom Checker

Overview

A conversational triage bot that asks follow-up questions based on initial symptom descriptions, then provides possible causes and recommended next steps (self-care vs. doctor).

Primary Use Cases

• Users unsure if they need medical attention
• Preliminary screening before telehealth visits
• Educational tool for basic triage

Key Features

• Chat UI with conditional flows (decision tree + generative fallbacks)
• Symptom ontology to map user input to standard terms (ICD-10)
• Follow-up question generator using LLM prompts
• Risk scoring based on severity/duration
• Report PDF summarizing conversation + suggested actions

Tech Stack

• Frontend: React + TypeScript + Tailwind
• Backend: FastAPI (Python) + Go service for risk scoring
• AI Models:
  • Classification: cambridgeltl/SapBERT-ICD10 fine-tuned for symptoms
  • Q-generation & advice: google/flan-t5-large
• DB: PostgreSQL (conversation logs, user feedback)

Architecture

1. User input → intent/symptom classification → map to ontology.