Zektor

Creates a copy-on-write clone of a PostgreSQL instance: it has the source's data from the moment of branching, and the two diverge independently afterwards. THIS COSTS MONEY — a branch is billed hourly like any instance: on its plan, and on what it writes itself above that plan's included storage at €0.15/GB/month, while it is a branch. Once promoted, it is billed on its full size. Pick a smaller plan than the source when the branch is for a test. Takes about a minute: the clone itself is instant, but a second PostgreSQL has to start. Poll get_instance until status is "active". Branching is refused on instances not using branch-capable storage.

RETURNS A LIVE CREDENTIAL. Creates a PostgreSQL role that stops working by itself after ttl_seconds and is then removed. This is the way to connect to a PostgreSQL instance: passwords are never stored, so get_connection cannot return one. The expiry is enforced by PostgreSQL itself, so the credential dies on time whether or not anything else is running. The password is returned once and exists nowhere else. Treat it as a secret and do not repeat it anywhere it would be stored or logged.

Provisions a database or cache. THIS COSTS MONEY — the plan is billed monthly from creation until the instance is deleted. Storage is not a parameter: it starts at the plan's included storage; grow it later with resize_storage. Creating runs in the background, so this returns an instance id and an actionId immediately. Poll get_action with the actionId until it reports "Success" (the instance is then "active") or "Failure".

Creates a PostgreSQL database for the next 60 minutes, with no account or token. Returns a connection string and a claim URL. The connection stops working after connectUntil; the data is deleted after claimUntil unless the user opens the claim URL, signs up and adds a payment method, which moves the data into a database of their own. Give the claim URL to the user: it is shown once. Limits: 200 MB, 3 connections, 30-second statements, no superuser; trusted extensions such as pgcrypto work. At most 3 trials per network per day.

PERMANENTLY DESTROYS a branch and anything written to it since it was created. The source database is untouched. Only works on branches — it refuses the cluster's default database, which has to be removed with delete_instance instead.

How the background work started by create_instance, scale_instance or resize_storage is going, by the actionId they return. status is "Running", "Success" or "Failure"; on Failure, errorMessage says why. While it is Running, call this again every few seconds.

RETURNS A LIVE CREDENTIAL for caches: the connection string includes the password. Treat the result as a secret and do not repeat it anywhere it would be stored or logged. For PostgreSQL there is no password to return — one is shown only when a role is created, and the server does not keep it — so the result carries host, port and database. Use create_ephemeral_role to obtain a usable credential.

Full detail for a single instance, including storage and autoscaling where it applies. Does not include credentials — use get_connection for those.

Size, usage, included storage, what is billed above the plan (net €/month), and automatic growth for a PostgreSQL instance.

The instance's cluster: the default database and every branch forked from it. `isDefault` marks the one promotion would replace.

Every database and cache on the account, with id, name, status, engine and plan.

The plans an instance can be created on or scaled to, with price, engine, memory, storage and the engine versions each one can run. Call this before create_instance or scale_instance rather than guessing a plan name.

Regions an instance can be created in. `available` is false where there is no capacity right now; creating there will be refused.

Promotes a branch to be its cluster's default database; the previous default becomes a branch. Nothing is deleted and no data moves — this changes which instance the cluster treats as the original. Connection details do not follow the promotion: each instance keeps its own host, port and credentials.

Grows the storage of a PostgreSQL instance to size_gb (larger than now, at most 1000 GB). Storage only grows: this can't be undone, not even by moving to a smaller plan. Each GB above the plan's included storage costs €0.15/month net, prorated by the hour; get_storage shows the current figures.

Changes the plan an instance runs on, which changes what it costs. Scaling DOWN reduces memory; storage stays (it never shrinks) and GB above the smaller plan are billed at €0.15/GB/month. If the instance is using more memory than the smaller plan provides, that becomes a problem during the move. Check current usage with get_instance first.

Turns automatic storage growth on or off for a PostgreSQL instance and sets its limit. When on, at 80% full storage grows to the larger of +5 GB or 70% full, never above limit_gb, which is required to turn it on; each grow is permanent and billed like resize_storage. A limit can be raised or lowered but not cleared. Only the fields you pass are sent, so changing one leaves the other alone.

Branches delete themselves after a while, so a forgotten one does not keep costing money. This extends that deadline, or removes it entirely. Removing it is a real decision, not a convenience: a kept branch holds its source's old blocks, which count against the source's storage. When that fills, growing it (by hand or automatically) is what costs, which is not obvious from looking at the branch.

Reports which Zektor account the current token belongs to, the API it is talking to, and whether the token came from the environment or the config file.