Skip to content

Latest commit

 

History

History
161 lines (110 loc) · 5 KB

railway-setup.md

File metadata and controls

161 lines (110 loc) · 5 KB

Railway Deployment Setup Guide

Overview

This guide explains how to deploy InterviewLab to Railway with multiple services:

  • API Service: FastAPI backend
  • Agent Service: LiveKit agent (optional, can run separately)

Prerequisites

  1. Railway account: railway.app
  2. GitHub repository connected
  3. Railway CLI installed: npm i -g @railway/cli

Step 1: Create Railway Project

railway login
railway init

Step 2: Add Services

PostgreSQL Database

railway add postgresql

Railway automatically sets DATABASE_URL environment variable.

Redis Cache

railway add redis

Railway automatically sets REDIS_URL environment variable.

Step 3: Deploy API Service

Option A: Using Railway Dashboard

  1. Go to Railway dashboard
  2. Click "New Service" → "GitHub Repo"
  3. Select your repository
  4. Railway will detect railway.json automatically
  5. Set environment variables (see below)

Option B: Using Railway CLI

railway link
railway up

Step 4: Configure Environment Variables

In Railway dashboard → Variables, set:

Variable Description Example
DATABASE_URL Auto-set by PostgreSQL service postgresql://...
REDIS_URL Auto-set by Redis service redis://...
SECRET_KEY JWT secret key Generate: openssl rand -hex 32
OPENAI_API_KEY OpenAI API key sk-...
LIVEKIT_URL LiveKit server URL wss://your-project.livekit.cloud
LIVEKIT_API_KEY LiveKit API key ...
LIVEKIT_API_SECRET LiveKit API secret ...
ENVIRONMENT Environment name production
LOG_LEVEL Logging level INFO
CORS_ORIGINS Allowed origins https://your-frontend.vercel.app

Step 5: Deploy Agent Service (Optional)

The agent can run on Railway or separately.

On Railway:

  1. Create new service: railway service create interview-agent (or name it agent)
  2. Link to same project: railway link
  3. IMPORTANT: Configure the service manually in Railway dashboard:
    • Go to Railway dashboard → Agent service → Settings → Deploy
    • Set Start Command to: python -m src.agents.interview_agent start
    • Disable Healthcheck: Go to Settings → Healthcheck and disable it (LiveKit agents don't expose HTTP endpoints)
    • The service uses Dockerfile.agent which doesn't include a healthcheck
    • This ensures it uses the agent instead of the FastAPI server
  4. Set same environment variables as API (especially LiveKit credentials)
  5. Deploy: railway up

Note: Railway may not automatically detect railway-agent.json. Always verify:

  • The start command in the dashboard matches the agent command above
  • Healthcheck is disabled (agents connect via WebSocket, not HTTP)

Separate Deployment:

Deploy agent separately (e.g., on a VPS) for better isolation.

Step 6: Run Database Migrations

Migrations run automatically on startup (see railway.json startCommand).

To run manually:

railway run alembic upgrade head

Step 7: Configure Domain

  1. Railway dashboard → Settings → Generate Domain
  2. Note the domain: your-app.railway.app
  3. Update frontend NEXT_PUBLIC_API_URL to point to this domain

Service Configuration Files

railway.json (API Service)

  • Builds from Dockerfile
  • Runs migrations on startup
  • Starts uvicorn server
  • Health check on /health

railway-agent.json (Agent Service)

  • Builds from same Dockerfile
  • Starts LiveKit agent
  • Requires same environment variables

Resource Limits

Default Railway configuration:

  • CPU: 1.0 vCPU
  • Memory: 2GB
  • Disk: 10GB

Adjust in Railway dashboard → Settings → Resources.

Monitoring

  • Logs: Railway dashboard → Logs
  • Metrics: Railway dashboard → Metrics
  • Health: https://your-app.railway.app/health

Troubleshooting

Issue Solution
Build fails Check Dockerfile, verify dependencies
Database connection error Verify DATABASE_URL is set
Agent won't connect Check LiveKit credentials
High memory usage Increase memory limit or optimize code
Port binding error Ensure using $PORT environment variable

Next Steps