Deploy and Host FSRPPO Trading Bot with Railway

Deploy a multi-account FXCM trading dashboard with a Python trading engine, shared model catalog and a private ForexConnect gateway. Each account has its own settings, SQLite history, FXCM session and active model. The template includes the experimental model catalog and starts with trading paused.

About Hosting FSRPPO Trading Bot

Railway runs the web app and the ForexConnect gateway as two services. The web app serves the dashboard, API and trading engines; its persistent volume at /app/data holds account databases and the shared models. The gateway is reachable only through Railway's private network. A generated password protects the public dashboard. The first account starts in Demo without FXCM credentials or an active model. Connect your own account and choose a compatible model in the dashboard before enabling trading.

Set PORT=8000 on the web service and PORT=5000 on the gateway. The web service reaches the gateway through its private Railway domain on port 5000.

Why Deploy

Use this template to start with isolated account data, a private FXCM gateway, and a shared model catalog without copying anyone else's credentials.

Common Use Cases

Test FXCM Demo strategies and compare models with backtests.
Operate multiple FXCM accounts with separate engines and histories.
Inspect FSRPPO decisions, positions and trading activity.
Optionally sync results to your own Supabase project or send Telegram signals.

Dependencies for Hosting

Deployment Dependencies

An FXCM Demo account for live broker data and execution; Real is optional.
Railway persistent storage for SQLite and models.
A ForexConnect gateway built for linux/amd64 and Python 3.7.
Optional Supabase project and Telegram bot for extra integrations.

After Deployment

Find the generated WEB_PASSWORD on the web service and sign in as admin. Connect FXCM Demo, select a model compatible with the instrument and verify risk limits before starting the engine. Configure volume backups if your Railway plan supports them, or keep a restorable external copy. Test a restore. The included models are experimental; no performance or profit is guaranteed. Read the repository README for maintenance and account-isolation rules.