Home/Developers/Status

Service status

You will not find an uptime percentage or a past incident record here — because there is no production environment yet. The product is in Phase 0: every component is in development. This page will be fed by live health data the day we go to production.

Source of status: hand-updated phase information · not live measurement

Components

Every component, exactly where it is.

The states below describe the local development environment. "Development" means the component has code but promises nothing to anyone in production.

API (control plane)development · REST + SSE endpoints run locally
Consoledevelopment · four screens run locally
Metering enginedevelopment · per-second, in memory
Supply source adaptersdevelopment · keyless mock, real supply connection
Budget cap and invoice calculationdevelopment · cap and pausing work
Payments (PayTR / Stripe)not connected yet · first beta
Identity and sessionsnot connected yet · first beta
Persistent databasenot connected yet · data is in memory
Host agent (our own supply)Phase 1

The colours in this table: an orange dot means development, a grey dot means not started. A green dot in this table will be used only for a component running healthily in production — today none of them is in that state. A green dot on the other pages of this site means "written within Phase 0 and running in local development", not production health.

How will this page work in production?

Once we are in production this list will be fed by health probes rather than by hand: each component's response time measured from outside, its error rate and the time of the last check. An uptime percentage will be published only after we start measuring, and with the measurement window written next to it — we will not publish a number calculated backwards.

Why is there no incident record at all?

Because there is no system carrying customer traffic yet; a system that does not exist cannot have an outage. When the first real outage happens it will be published here with its date, its duration, the component affected and its root cause — the accurate record, not the flattering one.

Incident reporting

What will we do when there is an outage?

Let the policy be written now, so nothing is up for debate on the day of the first incident. The flow below takes effect when we go to production.

Detection and announcement

If a component fails its health check, the status here changes first, then a warning banner appears in the console. Affected users also get an email. We announce the incident even when we do not know the cause; "we are investigating" is a notification too.

Effect on your invoice

While the outage lasts, if your pod is failing its health check those seconds are not billed in the first place — you do not have to file anything, the metering engine writes those seconds to a separate counter. This is what "you don't pay for seconds that don't work" means on the day of an outage.

Closure and root cause

After the incident is closed, a closing note lands here: what happened, how long it lasted, who was affected, what changed to stop it happening again. A text that fixes the system, not one that looks for someone to blame.

SLA

A commitment is given at the tier where it can be kept.

In a marketplace the supply comes from the community and from providers; making the same promise for every machine would not be honest. That is why the SLA arrives together with a tier.

Marketplace tier today

No SLA, but a guarantee: you don't pay for seconds that don't work. If a node fails its health check, billing stops; moving the job to another host is the intended behaviour and will be wired with Temporal at the end of Phase 0.

Verified tier Phase 3

Audited data centres and contracted hosts. Uptime commitments, response times and compensation rules are given at this tier — this is the product enterprise sales sells. It does not exist yet; the contract text has not been written.

Support response planned

With the first beta the channel and the target response time will be written here. We are not making that promise today, because we have no on-call team to keep it.

If you ever see an uptime percentage, the measurement window and the source will be written next to it. We do not publish percentages without a source.

Planned maintenance

Maintenance is the opposite of a surprise.

The planned maintenance policy — it takes effect when we go to production:

  • Advance notice. Maintenance that affects users is announced on this page and by email at least 72 hours in advance; the window is written with its hours.
  • Low-traffic window. Maintenance is done in the interval where the workload is lowest across EU and TR hours.
  • Running pods are not touched. Control plane maintenance does not stop running pods; the console and the API may be briefly unresponsive while the compute keeps going.
  • Effect on your invoice. If maintenance causes your pod to fail its health check, those seconds are not billed. The maintenance is our job; the bill for it is not yours.
  • Emergency maintenance. For work that cannot wait, such as a security patch, the announcement comes as soon as possible rather than in advance, and the reason is written down.

This policy is a description of how we work, not a contract text. The binding texts are collected under Terms of Use.

Follow

When the status changes, you learn it here.

The changelog tells you what shipped, the security page tells you how it is protected. The promise stays the same: you don't pay for seconds that don't work, a hard budget cap, per-second billing.