Soramai · Status

All systems operational.

Live health for every sub-system that Soramai depends on. This page is updated manually for now — automated probe ingestion is on the roadmap. Subscribe to the RSS feed for incident notifications, or follow our update channel.

Components

Each component is independently scaled and monitored. Outages in one rarely cascade into another.

Marketing site

soramai.com — landing, docs, pricing, status pages. Hosted on Vercel.

Operational

Authentication

Sign-in (Google + Apple), token issuance, session refresh. Backed by Firebase Auth.

Operational

Dashboard

console.soramai.com — wallet, training launch, datasets, models, deployments.

Operational

Training pool

GPU pods that run fine-tuning jobs. A100 + H100 capacity across regions.

Operational

Inference endpoints

Autoscaling serverless workers serving deployed adapters via api.soramai.com.

Operational

Billing & payments

Stripe checkout, webhook ingest, wallet credits, refunds. Idempotent.

Operational

Object storage

Datasets, adapters, training logs, generated images. Cloudflare R2.

Operational

Email

Transactional receipts and notifications via Resend.

Operational

Service-level objectives

What 'operational' means in numbers. These are targets, not contractual guarantees — see pricing for enterprise SLAs.

MetricTarget
Dashboard availability99.5% per rolling 30-day window
Inference endpoint p99 latency< 800 ms (text), < 4 s (image)
Inference success rate> 99.9% excluding upstream model errors
Training job completion rate (platform-side)> 99% excluding user-config failures
Billing webhook processing< 60 s from Stripe event to wallet credit

Incident history

No incidents recorded since launch. When that changes, this section will show every public-facing incident with a root-cause summary, duration, and remediation.

No incidents in the last 90 days.

Reporting an issue

If you’re experiencing a problem that isn’t reflected on this page, email support@soramai.com or open a ticket from the support page. Include your account email, the affected resource (job id / deployment id / dataset id), and the time the issue started. We reply within one business day.