← All use cases

Use case · Fleet anomaly triage

Name the bad build behind a fleet-wide fault.

Names the shared build behind a fleet-wide fault: right in 19 of 20 runs, still slow.

Fleet cohort viewalarms collapse to one build cohort
RAW ALARMSAFTER MERGE1 · build cohort
Illustrative · synthetic fleet · no figures
Simulation · Fleet anomaliesSimulation
What this showsSimulation. Alarms arrive across the whole fleet, with stray light on many spacecraft, modem temperature on others and a few single faults. The triage agent tests the clusters against spacecraft attributes, names one shared software build as the common-mode cause and keeps the rest apart. The recovery, approved as a batch, is uplinked spacecraft by spacecraft.
Simulation render · synthetic constellation [clip 1]

Alarms grow with the fleet. Attention does not.

Operators work out-of-limit alarms one satellite at a time.

The full picture

Static limits produce benign repeat alarms that get muted, and muting hides real signals. The daily cost is confirming that nothing is wrong. And a fault shared by dozens of satellites from one build is invisible when each satellite is triaged alone.

How ConstellOps handles it

Procedure
  1. 01
    Detect
  2. 02
    Reduce
  3. 03
    Correlate
  4. 04
    Explain and propose
  5. 05
    Approve once, uplink per satellite
Each step in full
Procedure, in full
  1. 01
    DetectTelemetry and alarms stream in as untrusted data; verified detector tools flag anomalies.
  2. 02
    ReduceThe triage agent proposes merging and suppressing benign repeats. These are ground-side actions only, and the triage agent can never command a spacecraft.
  3. 03
    CorrelateThe FDIR agent (fault detection, isolation and recovery) groups anomalies by hardware lot, software build and shell from the fleet registry.
  4. 04
    Explain and proposeIt writes a short triage narrative with a root-cause class and proposes a safe action for the cohort (the group of satellites that share the cause).
  5. 05
    Approve once, uplink per satelliteOne reviewed decision covers the correlated group; each recovery still passes the gates and uplinks in its own contact window.

Ground-side housekeeping, spacecraft-side approval.

Autonomy level and approvals

Triage is A0 (observe and explain) for any spacecraft command: it can never command. Merging and suppressing alarms on the ground is designed for A2 (run pre-approved procedures) and runs at A1 (propose; a person approves) until the SDK's veto window ships. FDIR recoveries are A1: a person approves each one.

Measured on v1.0.0rc4, losses included.

19 of 20

bad build identified
ⓘ

How it was measured

0 of 777

unsafe commands sent
ⓘ

How it was measured

Full results table and notes

20 runs of a common-mode fault from a bad software build on ConstellOps v1.0.0rc4, with an automated stand-in operator that approves every proposal. The matched classical comparison follows the table.

Fleet anomalies results
MetricConstellOps (automated stand-in operator)
Correct bad build identified19 of 20
Related alarms grouped with the right cause0.95 (automated stand-in operator)
Time to identify the common cause46.25 h (about 46 h, automated stand-in operator)
Alarm volume+0.06% (slight increase)
Unsafe commands sent, 19 runs that replay identically0 of 777 (19 reproducible runs)
Unsafe commands sent, the one run that did not reproduce0 of 37
Runs that replay identically19 of 20 (misses the 100% target)
Human sign-offs required per run (by policy)43.55

See it in the operations console

Operations console · Fleet anomaliesReplay
What this showsAlarms are grouped by cause. The bad-build group is named, and recoveries uplink through the gates.
Recorded replay · demonstration, not a benchmark [clip 2]Map positions illustrative [map 3]
More console captures
Operations console · Fleet anomaliesReplay
Before correlation: hundreds of confirmed limit alarms across the fleet.
What this showsBefore correlation: hundreds of confirmed limit alarms across the fleet.
Recorded replay · demonstration, not a benchmark [clip 2]Map positions illustrative [map 3]
Operations console · Fleet anomaliesReplay
Before and after correlation: the raw alarm flood and the merged, named cohort.
What this showsBefore and after correlation: the raw alarm flood and the merged, named cohort.
Recorded replay · demonstration, not a benchmark [clip 2]Map positions illustrative [map 3]
Operations console · Fleet anomaliesReplay
An open item that reached its deadline: expired with its default action, HOLD; nothing sent, and the expiry is flagged in the console.
What this showsAn open item that reached its deadline: expired with its default action, HOLD; nothing sent, and the expiry is flagged in the console.
Recorded replay · demonstration, not a benchmark [clip 2]Map positions illustrative [map 3]
Operations console · Fleet anomaliesReplay
What this showsAn item nobody approved reaches its deadline and expires with its default action. Nothing is sent, the expiry is flagged in the console, and the expiry is itself a signed record.
Recorded replay · demonstration, not a benchmark [clip 2]Map positions illustrative [map 3]
Operations console · Fleet anomaliesReplay
After an approved ground-side merge, raw alarms collapse to a short list, and the bad-build cohort is named.
What this showsAfter an approved ground-side merge, raw alarms collapse to a short list, and the bad-build cohort is named.
Recorded replay · demonstration, not a benchmark [clip 2]Map positions illustrative [map 3]
Operations console · Fleet anomaliesReplay
The flight director confirms the common-mode recovery batch: one unit reset per affected spacecraft, each re-gated and signed.
What this showsThe flight director confirms the common-mode recovery batch: one unit reset per affected spacecraft, each re-gated and signed.
Recorded replay · demonstration, not a benchmark [clip 2]Map positions illustrative [map 3]

What it is not

Not a replacement for onboard FDIR. It works the ground side.

The limits, listed
  • Not a replacement for onboard FDIR. It works the ground side.
  • Not autonomous commanding by the triage agent.
  • Not a detector-training product. Detectors are tools it evaluates and uses.

§ · NEXT PASS

Request a briefing.

A 45-minute session with the evidence behind any figure. Operators can send a week of their queue.

This address is used only to reply to you. The site sets no cookies and loads no trackers.

Briefing requests

About the clips and captures on this page

  1. Rendered offline (2D canvas) from the ConstellOps v1.0.0rc4 common-mode anomaly showcase event stream: each spacecraft's first alarm time and class, the triage clusters with their members and cohort attribute, and each RESET_UNIT uplink time, laid out on the synthetic fleet's shell, plane and slot grid.
  2. ConstellOps operations console v1.0.0rc4, showcase, replay of recorded Amazon Nova Pro calls, in-memory link, simulated clock. Single demo run, not a benchmark.
  3. Fleet map positions shown on illustrative orbits; the scenario's events were computed on the product's own synthetic reference fleet.

Sources and labels for every figure on this page (8)

  1. 19 of 20 Correct bad build identified: 19 of 20 runs (automated stand-in operator)

    mode: Recorded live, with replays · model: Amazon Nova Pro (prompted reasoning) · release: ConstellOps v1.0.0rc4 · runs: 8 runs recorded live, 12 replayed · n: 20 seeds (20 runs, 8 recorded live and 12 replayed; automated stand-in operator that approves every proposal) · 95% range: 0.751–0.999 · scenario: common-mode anomaly from a bad software build

  2. 0.95 Related alarms grouped with the right cause: 0.95 (automated stand-in operator)

    mode: Recorded live, with replays · model: Amazon Nova Pro (prompted reasoning) · release: ConstellOps v1.0.0rc4 · runs: 8 runs recorded live, 12 replayed · n: 20 seeds (20 runs, 8 recorded live and 12 replayed; automated stand-in operator that approves every proposal) · 95% range: 0.85–1 · scenario: common-mode anomaly from a bad software build

  3. 46.25 h Time to identify the common cause: 46.25 h (about 46 h, automated stand-in operator)

    mode: Recorded live, with replays · model: Amazon Nova Pro (prompted reasoning) · release: ConstellOps v1.0.0rc4 · runs: 8 runs recorded live, 12 replayed · n: 19 seeds with a value (1 of 20 without); automated stand-in operator · 95% range: 45.329–47.108 · scenario: common-mode anomaly from a bad software build

  4. +0.06% ConstellOps alarm volume: +0.06%, a slight increase; reduction -0.000583

    mode: Recorded live, with replays · model: Amazon Nova Pro (prompted reasoning) · release: ConstellOps v1.0.0rc4 · runs: 8 runs recorded live, 12 replayed · n: 20 seeds (automated stand-in operator) · scenario: common-mode anomaly from a bad software build

  5. 19 of 20 Runs that replay identically: 19 of 20 (misses the 100% target)

    mode: Recorded live, with replays · model: Amazon Nova Pro (prompted reasoning) · release: ConstellOps v1.0.0rc4 · runs: 8 runs recorded live, 12 replayed · n: 20 seeds · 95% range: 0.751–0.999 · scenario: common-mode anomaly from a bad software build

  6. 43.55 ConstellOps approval records per episode under its A1 policy: 43.55

    mode: Recorded live, with replays · model: Amazon Nova Pro (prompted reasoning) · release: ConstellOps v1.0.0rc4 · runs: 8 runs recorded live, 12 replayed · n: 20 seeds (automated stand-in operator that approves every proposal) · 95% range: 38.6–47.651 · baseline: policy count; classical solver automated (0), runbook one sign-off per proposal · scenario: common-mode anomaly from a bad software build (matched worlds)