Skip to content

deployment.md

Latest commit

 

History

History
117 lines (92 loc) · 4.44 KB

deployment.md

File metadata and controls

117 lines (92 loc) · 4.44 KB

Cloud Deployment

Prerequisites for Cloud Deployment

1. Qdrant Cloud Setup (Required for both platforms)

# Sign up for Qdrant Cloud at https://cloud.qdrant.io
# Create a cluster and note down:
# - Cluster URL (e.g., https://xyz-abc-123.eu-central.aws.cloud.qdrant.io:6333)
# - API Key (generated in the cluster dashboard)

Hybrid Deployment (Recommended)

Deploy frontend to Vercel and backend to Railway for optimal performance:

Step 1: Deploy Backend to Railway

# Install Railway CLI
npm install -g @railway/cli

# Login to Railway
railway login

# Initialize Railway project (in project root)
railway init

# Deploy backend
railway up

Railway Environment Variables Configuration: Go to Railway Dashboard → Your Project → Variables and set:

Variable Value Description
QDRANT_URL https://your-cluster.eu-central.aws.cloud.qdrant.io:6333 Your Qdrant Cloud cluster URL
QDRANT_API_KEY your_qdrant_cloud_api_key API key from Qdrant Cloud dashboard
OPENAI_API_KEY sk-... Your OpenAI API key
QDRANT_COLLECTION_NAME student_loan_corpus Vector collection name
DATA_FOLDER data/ Path to data files in Railway
BACKEND_HOST 0.0.0.0 Allow external connections
BACKEND_PORT 8000 Port for FastAPI (Railway auto-detects)
VERCEL_FRONTEND_URL https://xxxxxx.vercel.app Vercel frontend URL for CORS
LOG_LEVEL INFO Logging verbosity

Generate Public Domain URL:

  1. Select deployed service in Railway dashboard
  2. Go to Settings → Networking → Public Networking → Generate Domain
  3. Set port to 8000 (FastAPI default)
  4. Copy the generated domain (e.g., https://xxxxx.railway.app)

Step 2: Deploy Frontend to Vercel

# Ensure correct Node.js version
source ~/.zshrc && nvm use default && nvm use default

# Deploy frontend to Vercel
vercel --yes

Vercel Environment Variables Configuration: Go to Vercel Dashboard → Your Project → Settings → Environment Variables and set:

Variable Value Environment Description
NEXT_PUBLIC_BACKEND_URL https://xxxxx.railway.app/ Production Railway backend URL (from Step 1), ensure it trails with a /
NEXT_PUBLIC_DEPLOYMENT_ENV vercel Production Enables browser storage adapter
NEXT_PUBLIC_QDRANT_COLLECTION_NAME student_loan_corpus Production Collection name for frontend

Important Notes:

  • Only NEXT_PUBLIC_* variables are accessible in the browser
  • Backend handles Qdrant/OpenAI connections, so frontend doesn't need those credentials
  • Redeploy frontend after setting environment variables

Benefits of Hybrid Deployment:

  • Vercel Frontend: Fast CDN, excellent Next.js support
  • Railway Backend: Large Python dependencies, WebSocket support, no size limits

Alternative: Full-Stack Railway Deployment

Deploy both frontend and backend to Railway as a single service:

# Deploy entire application to Railway
railway login
railway init
railway up

Railway Environment Variables Configuration (Full-Stack): Go to Railway Dashboard → Your Project → Variables and set:

Variable Value Description
QDRANT_URL https://your-cluster.eu-central.aws.cloud.qdrant.io:6333 Your Qdrant Cloud cluster URL
QDRANT_API_KEY your_qdrant_cloud_api_key API key from Qdrant Cloud dashboard
OPENAI_API_KEY sk-... Your OpenAI API key
QDRANT_COLLECTION_NAME student_loan_corpus Vector collection name
DATA_FOLDER backend/data/ Path to data files
BACKEND_HOST 0.0.0.0 Allow external connections
BACKEND_PORT 8000 Port for FastAPI
NEXT_PUBLIC_DEPLOYMENT_ENV railway Enables browser storage adapter
NEXT_PUBLIC_QDRANT_COLLECTION_NAME student_loan_corpus Collection name for frontend
LOG_LEVEL INFO Logging verbosity

Benefits of Full-Stack Railway:

  • Single deployment: Easier management, one domain
  • WebSocket support: Better real-time features
  • No size limits: Handle large Python dependencies
  • Automatic builds: Deploys on git push

Storage Adapters

The application automatically adapts storage based on deployment environment:

  • Local Development: FileSystemStorage (saves to experiments/ folder)
  • Cloud Deployment: BrowserStorage (saves to browser localStorage)