Dear Railway Support Team,

I am contacting you regarding a serious and abnormal increase in billing on my project “atendi-ai”, caused after deployment actions performed by Railway’s own AI assistant inside the platform environment.

Historical Usage:

• Historical monthly average: approximately US$17/month
• April 2026: approximately US$47
• May 7th, 2026: approximately US$12 consumed in a single day

The increase was completely outside the normal behavior of my infrastructure.

What Happened:

During troubleshooting and infrastructure management, Railway’s AI assistant performed simultaneous deployment and modification actions across multiple Evolution API servers without properly understanding the existing architecture.

This resulted in:

• duplicated service instances
• duplicated database and websocket connections
• replication of WhatsApp sessions
• abnormal CPU, memory and network consumption
• connection conflicts between servers

Affected services:

• SERVIDOREVOLUTION OFICIAL
• EVOLUTION03 SERVIDOR
• EVOLUTION04 SERVIDOR

The issue became critical because the assistant itself acknowledged the error and identified that multiple servers and duplicated connections were running simultaneously.

As soon as the problem was identified, EVOLUTION04 SERVIDOR was removed immediately in order to stop the duplicated consumption and stabilize the environment.

Financial Impact:

• Expected proportional usage for early May: approximately US$4
• Actual consumption registered: approximately US$12 in a single day
• Estimated abnormal excess caused by the incident: approximately US$8 for the affected day

Additionally, April’s unexpected increase also appears related to infrastructure duplication and abnormal scaling behavior.

Another important point:

During the analysis process, the AI assistant itself recommended migrating away from Railway to a VPS infrastructure due to the operational cost structure becoming financially inefficient for large-scale WhatsApp workloads.

Requests:

1. Detailed review of the affected billing period
2. Evaluation of refund or account credit related to the abnormal duplicated consumption
3. Technical clarification regarding the duplicated deployments executed by the AI assistant
4. Assistance and recommendations to prevent this issue from happening again

I am attaching screenshots showing:

• duplicated active services
• the assistant acknowledging the deployment mistake
• the corrective actions taken immediately afterward

I appreciate your attention and understanding, and I hope we can resolve this situation fairly.

Best regards,

Antony

Target Atende AI