Skip to main content
For the complete documentation index, see llms.txt.
The Ground-Track Evidence Inspector is a read-only page for reviewing the fixed mission-level ground-track reference, the recorded evaluations run against it, and the canonical crossing and ground-track-error (GTE) evidence behind each evaluation. It exists to make that evidence understandable — including when it is refused, incomplete, or only good for diagnosis — without ever treating a modeled prediction as flown, observed truth. Route: /orbits/ground-track/evidence?referenceId=<uuid>&evaluationId=<uuid>
This inspector is currently available only to Valar Space team accounts (email addresses on the @valar.space domain). It is not yet available more broadly to Operations users. If your account is outside that domain, the page renders nothing — see Access.

​
Overview

Everything on this page is recorded evidence, already computed and persisted elsewhere. The inspector never runs a calculation of its own, never edits the reference or an evaluation, and never authorizes, commands, or approves anything:
  • The reference is a fixed, immutable identity — the same 15/1 mission-level ground-track reference is reused, unchanged, by every routine replan against it.
  • An evaluation is one modeled trajectory segment checked against that reference. Every evaluation’s evidence is explicitly labeled predicted from a modeled trajectory segment — it is not observed, post-burn verification.
  • A crossing is one point where the modeled ground track is checked against the reference, carrying either an authoritative signed ground-track-error (GTE) value or a typed reason why one could not be produced.

​
Access

The inspector is gated by a single Reflag flag, ground-track-evidence-inspector. Today that flag’s audience rule restricts access to accounts on the @valar.space email domain — the Valar Space team, not the general Operations user base. There is no second flag, environment variable, or other setting that can unlock the page for anyone else. While your access is still being resolved, or once it resolves to denied, the page renders nothing at all — no partial layout, no placeholder, no flash of content. This is a deliberate default-deny: the inspector only ever shows evidence once access is confirmed.

​
Reaching the inspector

The inspector is not in the main navigation — you reach it through an exact link, not by browsing to it. Two URL parameters control what it shows: If referenceId is missing, the page shows only:
Open ground-track evidence from a campaign or an exact evaluation link.
If referenceId is present but evaluationId is not, the page shows the reference identity and its evaluation history, and the evaluation-detail area shows the same message above — the inspector never guesses and silently opens the newest evaluation for you. Select an evaluation from the history list to see its detail, or follow a link that already includes both parameters.

​
The fixed reference

The reference header is the one piece of context that never changes, no matter which evaluation you have selected below it — the 15/1 reference is reused unchanged across every routine replan. It shows: If the reference itself cannot be read, the page shows one of the error and access states below in place of this header.

​
Recorded evaluations

Below the reference header, a compact Recorded evaluations table lists every evaluation recorded against the current reference — for navigation and audit retrieval, never for starting a new evaluation. Each row shows the evaluation’s modeled segment, when it was recorded, its outcome, its qualification, and a truncated digest; selecting a row’s segment link opens that evaluation’s detail below, setting evaluationId on the URL without leaving the page. If the reference has no recorded evaluations yet, this area shows:
No recorded evaluations for this reference.

​
Evaluation detail and the predicted-vs-observed distinction

Once an exact evaluation is selected, the detail area shows its outcome (Recorded, Incomplete, or Refused), its qualification/refusal disposition (below), and this persistent, unconditional label:
Predicted from modeled trajectory segment. Not observed post-burn verification.
That label is shown for every evaluation regardless of its outcome or qualification — the inspector never lets a qualified or well-conditioned result read as flown truth. Alongside it, the detail area shows:

​
Qualification and refusal states

An evaluation’s disposition is one of the following. These are the exact states the inspector distinguishes — none of them is a command, a permission, a compliance verdict, or a claim of flight truth; they describe only what the recorded evidence supports:

​
Canonical crossing evidence and ground-track error (GTE)

For a Recorded evaluation, the inspector shows its canonical crossings — the audit source of record — as a chart and a table, both ordered by crossing ordinal. The chart is a bounded, already-paginated projection of the same crossings shown in the table; it never recalculates, fits, or interpolates a value the table does not already contain. Each crossing row shows its ordinal, its crossing epoch, a status, its signed GTE, its conditioning, and an expandable Evidence disclosure with the underlying candidate epoch, reference/modeled coordinates, phase and longitude diagnostics, and a raw-evidence digest. Phase and longitude values are always secondary diagnostics — they are never substituted for the GTE. The signed GTE — the ground-track error, in meters, positive on the left of the modeled direction of motion — is the authoritative metric, and it is shown only when the crossing’s status is Present. For every other status, the cell instead shows:
No authoritative GTE for this crossing.
alongside the typed reason for that status: The inspector never shows zero, a fit, a phase value, or a longitude value in place of a missing GTE — a non-Present status always shows the reason above instead of a scalar. If an evaluation has no canonical crossings recorded at all, this area shows:
No canonical crossing was recorded for this evaluation segment.

​
Loading, empty, and error states

Every region of the page — the reference header, the evaluation history, the evaluation detail, and the crossing evidence — resolves independently to one of these states: If you change which evaluation is selected while a request for the previous one is still in flight, the inspector discards that stale response — it never lets an older evaluation’s crossings appear under a newer selection.

​
What this page does not do

The inspector is read-only, end to end. Every action on the page is inspection, selecting a row, following provenance (a digest or source identifier shown for reference), or reading exported evidence — never any of the following:
  • No editing or reseeding the reference or an evaluation
  • No campaign, grid, or reference authoring
  • No submitting, planning, or scheduling a new evaluation
  • No maneuver planning, approval, or command
  • No declaring or recording compliance
Qualification and diagnosis-only labels describe what the recorded evidence supports for downstream use — they are not a permission, an approval, or a flight-truth verdict, and nothing on this page can grant one.

​
Evidence and certification status

The evidence this page shows is backed by an additive numerical verification and release-certificate process that runs in CI against the platform’s own persisted reference, evaluation, and crossing artifacts. That process establishes bounded numerical feasibility for the evidence you see — it is CI/evidence tooling, not a flight-readiness or command system, and it does not itself grant final release, campaign compliance, or observed post-burn verification. Where that certification remains non-final, the inspector’s own qualification and refusal states (above) are the accurate, current description of what a given evaluation supports — this page never states a final qualification or release beyond what those recorded states already say.