# railway

> Deploy and manage applications on railway.com. Use when the user asks to deploy, host, or ship an app to Railway, add databases or services to Railway, configure Railway projects, or troubleshoot Railway deployments.

- Author: Paul-Louis Pröve
- Repository: pietz/skills
- Version: 20260121171129
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-06
- Source: https://github.com/pietz/skills
- Web: https://mule.run/skillshub/@@pietz/skills~railway:20260121171129

---

---
name: railway
description: Deploy and manage applications on railway.com. Use when the user asks to deploy, host, or ship an app to Railway, add databases or services to Railway, configure Railway projects, or troubleshoot Railway deployments.
---

# Railway Deployment

## CLI Basics

The Railway CLI has built-in help for all commands:

```bash
railway --help
railway <command> --help
```

**Important:** The CLI has interactive menus that don't work in automated contexts. Always provide all required parameters explicitly to avoid prompts (e.g., `--workspace`, `--service`, `--name`).

## Common Commands

```bash
# Create project (always specify workspace to avoid prompt)
railway init --name "my-project" --workspace "Workspace Name"

# Deploy from a template (get ID from template URL on railway.com)
railway init --template <template-id>

# Add PostgreSQL database
railway add --database postgres

# Add empty service for your app
railway add --service api

# Link local directory to service
railway service link api

# Set environment variables (use references for database URLs)
railway variables --service api --set 'DATABASE_URL=${{Postgres.DATABASE_URL}}'

# Deploy
railway up

# Generate public domain
railway domain --service api

# View logs
railway logs --service api
railway logs --service api --build    # build logs

# SSH into running service
railway ssh --service api
```

Always bind your app to `0.0.0.0:$PORT` - Railway injects the PORT variable.

## Logging Setup (Python)

Railway expects JSON logs on stdout with a lowercase `level` field. Use `python-json-logger`:

```bash
uv add python-json-logger
```

```python
# logging_config.py
import logging
import sys
from pythonjsonlogger.json import JsonFormatter

class RailwayJsonFormatter(JsonFormatter):
    """JSON formatter with lowercase level names for Railway."""
    def add_fields(self, log_record, record, message_dict):
        super().add_fields(log_record, record, message_dict)
        log_record["level"] = record.levelname.lower()

def setup_logging():
    handler = logging.StreamHandler(sys.stdout)
    handler.setFormatter(RailwayJsonFormatter("{message}", style="{"))
    logging.basicConfig(handlers=[handler], level=logging.INFO, force=True)

    # Silence noisy third-party libraries
    for name in ("uvicorn", "uvicorn.error", "uvicorn.access", "httpx", "httpcore", "sqlalchemy"):
        logging.getLogger(name).setLevel(logging.CRITICAL)
```

Call `setup_logging()` at app startup before creating loggers.

### PostgreSQL/pgvector Log Levels

PostgreSQL INFO messages appear as errors in Railway by default. Fix with a custom start command in Settings:

```
/bin/sh -c "unset PGPORT; exec docker-entrypoint.sh postgres --port=5432 -c log_min_messages=warning -c log_statement=none 2>&1"
```