Konsulta — Smart Triage Widget

An AI symptom-intake widget. When a patient describes their symptoms in free text, Claude returns:

• Recommended department
• Urgency (routine / soon / urgent / emergency)
• 2–3 matching doctors + suggested appointment slots

Designed as a thin "middleware" layer that sits in front of an existing booking system (e.g. eKonsulta Clinic). It adds an intelligent intake step before the booking page without touching the backend.

Tech Stack

Area	Technology
Frontend	React (Vite)
Backend	Node.js + Express (REST API)
DB	MongoDB (Mongoose)
AI	Claude (claude-opus-4-8) via @anthropic-ai/sdk
Infra	GCP (target: Cloud Run / App Engine)

This is a dummy-data demo. It always runs: with no MongoDB connection it falls back to in-memory dummy data, and with no ANTHROPIC_API_KEY it falls back to rule-based mock triage.

How to Run

Run the backend and frontend in two separate terminals. The backend must be running first so the frontend's /api proxy can reach it.

Prerequisites: Node.js 18+ (npm). MongoDB and an Anthropic API key are optional — without them the app falls back to in-memory data + rule-based mock triage.

1. Backend (Terminal A) — http://localhost:4000

cd server
cp .env.example .env          # creates server/.env (NOT the repo root)
                              # optionally set ANTHROPIC_API_KEY, MONGODB_URI
npm install
npm run seed                  # (optional) load dummy doctors into MongoDB — see "Database" below
npm run dev                   # start the API server

Where does .env live? Because the cp above runs from inside server/, it creates server/.env. The backend's dotenv loads the .env from the directory the server runs in (server/), so a .env in the repo root is not used. Keep the env file at server/.env.

Restart after editing .env. npm run dev uses node --watch, which only watches imported source files — not .env (it's read by dotenv, not imported). After changing MONGODB_URI, ANTHROPIC_API_KEY, etc., stop the server (Ctrl+C) and run npm run dev again, or it keeps using the old values.

2. Frontend (Terminal B) — http://localhost:5173

cd client
npm install
npm run dev                   # start the Vite dev server, then open the URL it prints

Vite proxies /api/* to the backend on port 4000 (see client/vite.config.js), so no extra config is needed.

Production build (frontend)

cd client
npm run build                 # outputs static files to client/dist/
npm run preview               # preview the built site locally

Database (MongoDB)

The app works without any database (in-memory dummy data). To persist the dummy doctors and triage logs, point MONGODB_URI in server/.env at a MongoDB instance and run the seed once.

Option A — MongoDB Atlas (cloud, no local install)

1. Create a free M0 cluster at https://www.mongodb.com/cloud/atlas.
2. Database Access → add a user (remember the password).
3. Network Access → add your IP (or 0.0.0.0/0 for development).
4. Connect → Drivers → copy the connection string.
5. Put it in server/.env, replacing <password> and adding the database name /konsulta in the path (before the ?):

# server/.env  — note the "/konsulta" path segment
MONGODB_URI=mongodb+srv://<user>:<password>@cluster0.xxxxx.mongodb.net/konsulta?retryWrites=true&w=majority

⚠️ Common mistake: the database name goes in the path (.../mongodb.net/konsulta?...), not as a query parameter (...?konsulta=...). Without /konsulta the data lands in the default test database.

Option B — Local MongoDB

# macOS (Homebrew)
brew install mongodb-community
brew services start mongodb-community

# server/.env
MONGODB_URI=mongodb://localhost:27017/konsulta

Seed the dummy data

cd server
npm run seed     # → "[db] MongoDB connected" then "[seed] Loaded 16 doctors"

Once connected, npm run dev logs [db] MongoDB connected, and the API reads doctors from the doctors collection and writes triage logs to triagerequests. Without a reachable MONGODB_URI it silently falls back to in-memory data.

The "Demo mode (rule-based)" tag in the widget footer reflects the AI status (no ANTHROPIC_API_KEY), not the database. The DB works independently.

API

POST /api/triage

{ "symptoms": "Chest has felt tight for 3 days and I get short of breath climbing stairs", "locale": "en" }

Response:

{
  "recommendedDepartment": "Cardiology",
  "urgency": "urgent",
  "urgencyLabel": "Urgent — seek prompt care",
  "assessment": "Chest discomfort and shortness of breath worsening with exertion...",
  "redFlags": ["chest pain on exertion", "shortness of breath"],
  "doctors": [
    {
      "id": "doc-cardio-01",
      "name": "Dr. Maria Santos",
      "department": "Cardiology",
      "specialties": ["..."],
      "languages": ["Filipino", "English"],
      "suggestedSlots": ["2026-06-18 09:30", "2026-06-18 10:00"]
    }
  ],
  "source": "claude"
}

GET /api/doctors

Returns all dummy doctor/department data.

Structure

konsulta/
├── server/   Express REST API + MongoDB + Claude triage service
└── client/   React widget