MediQuery 🏥

MediQuery is a natural language to SQL engine for hospital analytics. Ask questions about hospital data in plain English — the app generates SQL using Groq LLM and returns results from a structured hospital database.

Live Demo: https://mediquery.vercel.app Backend API Docs: https://mediquery-api-uu2g.onrender.com/docs

---

How It Works

User Question (plain English)
        │
        ▼
FastAPI Backend (Render)
        │
        ▼
LangChain SQLDatabaseChain
        │
        ├── Groq LLM generates SQL query
        │
        ▼
SQLite Hospital Database
        │
        ▼
Results returned as JSON
        │
        ▼
Next.js Frontend renders table

---

Tech Stack

Layer	Tool
Frontend	Next.js + Tailwind CSS
Frontend Hosting	Vercel
Backend	FastAPI
Backend Hosting	Render
NL to SQL	LangChain SQLDatabaseChain
LLM	Groq (llama-3.1-8b-instant)
Database	SQLite
Containerisation	Docker

---

Database Schema

Hospital data warehouse with 5 tables:

dim_patient      — patient demographics (1,751 records)
dim_staff        — staff records (262 members)
dim_dept         — hospital departments (10 departments)
dim_bed          — bed assignments
fact_treatment   — treatment records (cost, LOS, rating, feedback)

---

Example Queries

"Show total treatment cost by department"
"How many patients are in each status category?"
"What is the average rating per department?"
"List top 5 patients by length of stay"

---

Project Structure

mediquery/
├── core/
│   └── chain.py          ← LangChain NL-to-SQL chain + DB stats
├── api/
│   └── routes.py         ← FastAPI route handlers
├── frontend/             ← Next.js app (deployed on Vercel)
│   ├── app/
│   │   ├── page.tsx      ← main page
│   │   └── layout.tsx    ← root layout
│   └── components/
├── main.py               ← FastAPI entry point
├── hospital_warehouse.db ← SQLite hospital database
├── Dockerfile
├── requirements.txt
└── .env.example

---

Run Locally

Prerequisites

• Python 3.11+
• Node.js 18+
• Groq API key → console.groq.com

Backend

git clone https://github.com/rohitsahayy/mediquery.git
cd mediquery

cp .env.example .env
# add GROQ_API_KEY to .env

pip install -r requirements.txt
uvicorn main:app --reload
# API running at http://localhost:8000
# Docs at http://localhost:8000/docs

Frontend

cd frontend
npm install

echo "NEXT_PUBLIC_API_URL=http://localhost:8000" > .env.local

npm run dev
# UI running at http://localhost:3000

Docker (backend only)

docker build -t mediquery .
docker run -p 8000:8000 --env-file .env mediquery

---

API

POST /api/query

// Request
{ "question": "Show total treatment cost by department" }

// Response
{
  "sql": "SELECT dd.Dept_Name, SUM(ft.Treatment_Cost) ...",
  "columns": ["Dept_Name", "Total_Cost"],
  "rows": [["Cardiology", 245000.0], ["Neurology", 198000.0]],
  "row_count": 10
}

GET /api/stats

{
  "total_patients": 1751,
  "departments": 10,
  "staff": 262,
  "treatment_records": 1751
}

GET /api/examples

{
  "examples": [
    "Show total treatment cost by department",
    "List top 5 patients by length of stay",
    "How many patients are in each status category?",
    "What is the average rating per department?"
  ]
}

Full interactive docs: https://mediquery-api-uu2g.onrender.com/docs

---

Environment Variables

Backend .env:

Variable	Description
GROQ_API_KEY	Groq API key for LLM inference

Frontend .env.local:

Variable	Description
NEXT_PUBLIC_API_URL	Backend API base URL

---

Deployment

Service	Platform	URL
Frontend	Vercel	https://mediquery.vercel.app
Backend	Render	https://mediquery-api-uu2g.onrender.com

Auto-deploys on every push to main.

Note: Backend is on Render free tier — first request after inactivity may take 30–50 seconds to wake up.