Skip to content

Latest commit

 

History

History
140 lines (97 loc) · 3.18 KB

DEPLOYMENT.md

File metadata and controls

140 lines (97 loc) · 3.18 KB

Railway Deployment Guide

This guide covers deploying the pneumonia detection web application to Railway.

Prerequisites

  • GitHub repository with the application code
  • Railway account (https://railway.app)
  • Model API endpoint URL (from the model team)

Deployment Steps

1. Connect Repository

  1. Log in to Railway
  2. Click "New Project" → "Deploy from GitHub repo"
  3. Select your repository
  4. Railway will auto-detect the Python project

2. Configure Environment Variables

In Railway dashboard, go to your service → Variables tab and add:

Required:

MODEL_API_URL=https://your-model-api.railway.app/predict
SECRET_KEY=<generate-a-random-32-char-string>

Optional:

MODEL_API_HEALTH_URL=https://your-model-api.railway.app/health
MODEL_API_KEY=<if-authentication-required>
API_TIMEOUT_SECONDS=10
MAX_FILE_SIZE_MB=10
DEBUG=False

To generate a secret key:

python -c "import secrets; print(secrets.token_hex(16))"

3. Deploy

  1. Railway automatically deploys when you push to main branch
  2. Or manually trigger deploy from the dashboard
  3. Wait for build to complete (usually 1-2 minutes)

4. Verify Deployment

  1. Click the generated URL in Railway dashboard
  2. Check the health endpoint: https://your-app.railway.app/health
  3. Expected response: 
    {"status": "healthy", "model_api": "connected"}

5. Test the Application

  1. Open the app URL in browser
  2. Upload a test X-ray image
  3. Verify you receive a prediction result
  4. Check error handling with invalid files

Configuration Files

Railway uses these files:

  • Procfile: web: gunicorn app.main:app
  • runtime.txt: python-3.10.12
  • requirements.txt: Python dependencies

Monitoring

View Logs

  1. Go to Railway dashboard → your service
  2. Click "Logs" tab
  3. Filter by deployment or view live logs

Health Checks

Railway automatically monitors your app. The /health endpoint provides:

  • Overall application status
  • Model API connectivity status

Updating the Application

  1. Push changes to main branch
  2. Railway automatically redeploys
  3. Zero-downtime deployment

Rollback

If a deployment has issues:

  1. Go to Railway dashboard → Deployments
  2. Find the previous working deployment
  3. Click "Rollback" to restore

Custom Domain (Optional)

  1. Go to Settings → Domains
  2. Add your custom domain
  3. Configure DNS CNAME record as instructed

Troubleshooting

App won't start

  • Check Procfile syntax
  • Verify requirements.txt is complete
  • Check logs for import errors

Model API connection fails

  • Verify MODEL_API_URL is correct
  • Check if model API is deployed and running
  • Try the health check endpoint directly

Slow response times

  • Model inference can take a few seconds
  • First request after idle may be slower (cold start)
  • Consider keeping the API warm

Memory issues

  • Railway free tier has limits
  • Optimize image handling to reduce memory
  • Consider upgrading plan for production use

Environment Comparison

Setting Development Production
DEBUG True False
SECRET_KEY dev-key Random secure key
MODEL_API_URL localhost or test Production API URL