Skip to content

Observing accounts

An observing account is the practical handle you submit observations against. It bundles queue access on one or more telescopes into a single unit that you (or your research group, or your class) can use without thinking about which queue to hit for each request. An account reaches at most one queue per telescope — so picking an account is picking the lane on every telescope it spans — and the scheduler picks an eligible telescope from whatever the account exposes.

If you’re new to the access model: telescope owners issue queue access grants to entities (users or organizations). Those entities then bundle their grants into observing accounts, and the grants on the account decide who can actually use it, and in what order.

This page is the observer-side view. For the owner-side mechanics, see Telescope access.


An ObservingAccount carries:

  • Owner — the entity (user or organization) the account belongs to. The owner controls who can submit through it.
  • Queue access — the ObservingQueueAccessGrant records that connect the account to telescope queues, at most one per telescope. These determine which telescopes are reachable from this account and in which lane on each.
  • Observing grants — the ordered list of records (OwnerObservingGrant, GroupObservingGrant, MemberObservingGrant, internal buckets, external delegations) that say who can submit observations through the account and who goes first under contention. See Observing grants and Priority levels.
  • Quotas — zero or more ObservingQuota records that throttle how much the account can spend within a period. See Quotas.
  • credits_used — running total of credits consumed by the account so far. Each transaction is recorded as an ObservingAccountTxn linked to the originating observation task.
  • Level (order) — on a root account, where it ranks among your other root accounts that share a queue. See Priority levels.
  • Subaccounts — accounts form a tree: buckets you carve out for yourself, and delegations to other entities.

You’ll see your accounts on the Observing Accounts page of your user profile (/u/<your-slug>/observing-accounts) or of an organization you manage (/orgs/<slug>/observing-accounts) — split into Your accounts (root accounts you created from your own queue access), Buckets (subaccounts you carved out of your own accounts), Delegated to you (subaccounts another entity sponsors) and Archived — and as the very first step of the observation editor (an observation can’t be published until you’ve picked at least one funding grant). The dashboard’s /dashboard/observing-access page lists the observing grants you hold, the telescopes they reach, and the observation types you can submit.


Created by a user or organization that holds queue access grants directly.

  • Funded through queue access grants, at most one per telescope.
  • Distribute access downstream through observing grants, buckets, and delegations.
  • Carry a level among your other root accounts.
  • Accounts you set up from your own telescope access are root accounts. The First Light account every new user gets is not: it is a delegated subaccount.

Created under one of your own accounts via an internal grant, to slice your access into units you manage separately — a research program, a class, an outreach effort.

  • Owned by the same entity as the parent; optionally managed day-to-day by one of the organization’s groups.
  • Have their own grants, quotas, ledger, and further subaccounts.
  • Reach exactly the lanes the parent reaches; they redistribute access, they never add it.
  • Ranked by the parent among the parent’s grants.

Created under an account via a delegated observing grant to another entity. Used to give an outside collaborator time from your allocation without giving them visibility into the parent.

  • Owned and fully managed by the recipient. When the recipient is a user, the subaccount comes with an owner grant so they can submit right away; an organization recipient adds its own grants.
  • Reach exactly the lanes the parent reaches.
  • Ranked by the sponsor among the sponsor’s grants; the recipient sees that standing but cannot change it.
  • The sponsor sees the credits the branch uses, not its observations, unless the sponsor’s account turns on sponsor view.

Priority inside your slice of a queue is expressed as levels on two kinds of lists, and every list starts as a single level, so everything ties until you decide otherwise. Ties are good: the more observations at the same level, the more freedom the scheduler has to pack the night.

  • Your root accounts. The Your accounts list can be split into levels. Levels only matter between accounts that reference the same queue — an organization running a “research” account and an “outreach” account against the same grant can put research first — and an account that spans several telescopes sits at the same level on each of them.
  • The grants on an account. Each account’s grant list can be split into levels: a faculty group at level 1, individual thesis students at level 2, the whole organization at level 3, and a partner’s delegation at level 4. Buckets and delegations are ranked in the same list as rosters. Inside a bucket or a delegated subaccount, that account’s own managers rank its grants the same way.

Rules worth knowing:

  • Levels are entirely local to your tree: they never change how much telescope time you get relative to other grantees (the telescope owner’s queue configuration decides that — see Queue prioritization), and other entities never see them, except that a delegated recipient sees the level its own subaccount holds in the sponsor’s list.
  • Nothing below an edge can move work above that edge’s level. A delegated recipient can order everything inside their subaccount, but the subaccount’s standing among your grants is yours.
  • An observation funded through several grants on one account is funded through the best-ranked one. Giving someone a personal member grant at a higher level is how you promote them above the org-wide default, provided their observations are funded through that grant.
  • New grants (and new root accounts) join the top level unless you choose another; move them in the level list afterwards.
  • Among equals, the observation that has been waiting longest goes first.

An observing grant says “this entity is allowed to submit through this account” — or, for buckets and delegations, “this branch hangs here.” Every grant has a level in the account’s list, tracks its own credits_used, and can carry its own quotas. Grants have several flavors:

Gives the account’s owner the right to submit. It is not added automatically to a root account: the New account dialog for a personal account offers it (“Also grant myself access to this account”, on by default), and a delegated subaccount given to a user gets one. If the owner is a user, that user can submit. If the owner is an organization, every member of the organization can submit (subject to organizational RBAC).

Assigned to an individual user. Lets a single member of an organization submit through an account they don’t own — useful when you want fine-grained “who can submit” control inside a larger organization, or when you want to rank one person above the organization-wide grant.

Assigned to a group. Every member of the group can submit.

A grant to yourself that creates a bucket — a subaccount owned by the same entity, with its own grants, quotas, and ledger, optionally managed by one of your groups. Use it when a slice of your access needs several rosters, its own quota pool, or its own managers. The bucket’s level among your grants is set here, on the grant.

A grant of authority to another entity (API grant type externally_managed) that creates a subaccount they own under your account — not just to submit through yours. You can name a user or an organization you administer directly, or invite someone by email; an emailed invitation creates the subaccount when it is accepted and expires after 14 days. The recipient consumes time against your account but tracks it independently and manages their own people (consortium / partnership workflow). Its level among your grants is set here, on the grant; everything inside the subaccount is the recipient’s.


A quota is the mechanism Skynet uses to throttle credit consumption. A quota is essentially a rule of the form “don’t spend more than max_credits within a period_type window.”

The period_type is drawn from a small set:

period_type What it means
lifetime No window — the cap is a fixed total over the lifetime of the account or grant, and it never resets on its own
day Calendar day, midnight to midnight in the quota’s timezone
week Calendar week, Monday 00:00 to the next Monday 00:00 in the quota’s timezone
month Calendar month, from the 1st at 00:00 in the quota’s timezone
night “Observing night”: 24 hours starting at starts_at_local_time (12:00 by default) in the quota’s timezone, so observations early and late in the night roll up to the same period

Each quota also carries:

  • max_credits — the cap for the period.
  • timezone — the IANA timezone the period boundaries are computed in (default UTC).
  • starts_at_local_time — the local time a night period starts (default 12:00); ignored by the other periods. anchor_at is accepted but not used yet.
  • enabled — quotas can be disabled without being deleted (so a policy that’s temporarily relaxed doesn’t lose its configuration).

Both accounts and grants can carry multiple quotas at once. When they do, every enabled quota must have credits left — the scheduler stops placing work as soon as any one of them is used up for its current period.

Useful combinations:

  • Lifetime + weekly — “10 000 credits total, but no more than 500 in any one week.” The lifetime quota caps the program’s total reach; the weekly quota smooths consumption.
  • Daily + lifetime — a research group with a steady budget (“400/day, max 50 000 lifetime”).
  • Nightly — for classroom workflows where the policy is “everyone gets the same time per observing night.”

There are two places quotas can live:

  • On an account (an ObservingAccountQuota, keyed by accountUid) — throttles the collective consumption of all grants under that account, including everything spent in its buckets and delegated subaccounts.
  • On a grant (an ObservingGrantQuota) — throttles a single grantee’s consumption — useful when an account has many grantees and you want per-grantee caps in addition to the account-wide cap. On a bucket or delegation grant it caps the whole branch.

Both layers compose. If an account has a 500/week quota and a specific member grant under it has a 100/week quota, that member caps out at 100/week and the broader pool of members caps out at 500/week collectively.

A quota blocks once it has no credits left in its current period (credits are charged when tasks complete, so work already under way when the cap is reached still finishes and can take usage past the cap). From then on the scheduler stops placing tasks funded through any grant whose chain includes that quota — the grant’s own quotas, its account’s, and those of every account and delegation grant above it. If the observation also has a funding grant with headroom left, it keeps scheduling through that one. Otherwise the observation stalls — it does not error out — and resumes once the quota period rolls over (a lifetime quota never rolls over; it has to be raised) or a new grant is attached. Quotas never change an observation’s priority; they only pause it.

If your observation isn’t making progress, the first thing to check is whether one of your quotas has filled up. The observation’s overview says so when its funding is over quota; on the account and grant pages, compare the credits used in each quota’s current period against its cap.


Submitting an observation through an account

Section titled “Submitting an observation through an account”

In the observation editor, the accounts step is the first thing you do. It lists the grants you can fund the observation through, on accounts owned by the observation’s owner, with each account’s queue on every telescope it reaches and the account and grant levels. Picking a grant determines the lane on every telescope its account spans; if you pick grants on several accounts that reach the same telescope, the scheduler uses the higher lane there. The observation is funded through the grants you pick, and on each telescope the scheduler charges the best-ranked one that still has quota.

For the editor-side detail, see Editor → Accounts step.


Every new user gets a first-light observing account the moment they sign up. From the user’s perspective it’s an ordinary observing account with a quota — they can immediately create observations and projects against it, without needing any specific telescope grant to be issued for them.

Mechanically:

  1. Skynet maintains a single parent first-light observing account funded by Skynet-owned queue access grants across the telescopes that participate in the first-light program.
  2. When the new user’s account is created, Skynet automatically issues a delegated observing grant from that parent account to the new user, creating a subaccount under it with an owner grant for the user.
  3. The user’s subaccount inherits the parent’s queue access (so it reaches the same telescopes). The delegated grant carries a weekly quota, set by Skynet, that caps how much the new user can consume per week. First Light time is for the new user alone: the delegated grant does not allow delegation, so they cannot delegate their First Light account (or anything under it) to someone else.

The first-light account is just an observing account configured this way; no special model fields, no first-light flag. The same machinery (delegated grants, subaccounts, quotas, levels) is what classrooms and organizations use for their own sub-allocations, just with different parents and policies.


Credit accounting and how observations get charged

Section titled “Credit accounting and how observations get charged”

A few things worth knowing up front:

  • Creating an observation costs nothing. You can create as many observations as you want without consuming any credit. The editor never blocks creating one for quota reasons.
  • Credits are tallied as tasks complete. Each task is funded through the one grant the scheduler picked for it. When the telescope finishes the task, it is charged its planned duration in seconds (for optical imaging, scaled by the imager’s efficiency factor and rounded up). An ObservingAccountTxn is recorded against the funding grant and its account, and the same amount is copied up every delegation or bucket edge to the root account, so each level’s credits_used and ledger include the spending below it.
  • Throttling is quota-driven. Accounts and grants don’t carry a single “limit” field — they carry zero or more quotas, each defining a (period, max_credits) ceiling. The scheduler stops placing work when any enabled quota on the funding chain is used up; tasks already running finish, and scheduling resumes when the quota period rolls over or the quota is raised.

Practical consequence: a “stalled” observation is much more likely to be a quota state than a scheduling bug. Inspect each of:

  1. The quotas on the account and their use in the current period.
  2. The quotas on each funding grant and their use in the current period.
  3. Quotas above your account: on a bucket or delegated subaccount (including First Light), the quotas on the grant that connects it to its parent and on the parent accounts apply too.

Any enabled quota on that chain with nothing left in its current period is what’s holding the observation up.


The full model lives across these schemas:

  • ObservingAccount — the account itself.
  • ObservingQuota — quota limits, on an account or a grant.
  • ObservingAccountTxn — transaction history.
  • ObservingGrant and subclasses (OwnerObservingGrant, MemberObservingGrant, GroupObservingGrant, and the delegated family DelegatedObservingGrant → InternalObservingGrant, ExternallyManagedObservingGrant) — the grant family.
  • ExternallyManagedObservingGrantInvitation — an emailed offer of a delegated grant.