LoopgraphPaper
Preview mode — rich sample data is shown here so you can see the Hermes Brain flow before connecting your own project.
Operate

Learning evidence

See whether Hermes is making better routing decisions, which evidence packet each decision acknowledged, where humans corrected it, and whether the selected loops produced observed outcomes.

Open loop definitions
Illustrative hosted preview. These records demonstrate the complete operating cycle and are never written into an installed workspace. Local Loopgraph shows only the opportunities, decisions, measurements, and value created from your own Hermes events.
Route decisions
86
78 committed
Evidence bound
95%
82 acknowledged · 3 stale rejected
Routing evaluation
94%
45 passed · 3 failed
Human corrections
5
Accountable route feedback
Outcomes
4
Completed evaluations
Observed
2
Sufficient real evidence
Incomplete
1
Hermes needs more evidence
Failed jobs
1
0 dead-letter

Recent Hermes decisions

Every row distinguishes evidence-bound decisions from legacy or stale submissions. A stale acknowledgement is rejected before it can create loop work.

ReceivedEventDecisionSelected loopsEvidenceResult
Jul 28, 5:48 PMevt_product_account_inactivityRouteproduct_activation_recoveryAcknowledgedCommitted
Jul 28, 5:31 PMevt_support_enterprise_riskRequest HumanNo loop selectedAcknowledgedCommitted
Jul 28, 5:08 PMevt_ads_qualified_pipeline_dropRoutemarketing_adsUnacknowledgedStale evidence rejected

Learning by loop

Selection volume is shown beside golden-event quality and accountable corrections, so frequently chosen loops are not mistaken for accurate loops.

LoopSelectedEvaluatedPass rateCorrections
product_activation_recovery211894%1
support_priority_triage191694%2
marketing_ads14989%1
marketing_content_creation75100%1

Measurement pipeline

Each row shows the exact source, collection state, and most recent evidence for one loop metric.

Loop / metricBindingCollectionLatest evidenceState
product_activation_recovery
qualified_activation_rate · Primary
posthog-production
product.analytics.read
Every 1h
14 complete · 1 pending · 0 failed
68.4 %
Jul 28, 5:00 PM
Verified
support_priority_triage
urgent_first_response_minutes · Primary
support-production
support.tickets.read
Every 30m
28 complete · 0 pending · 0 failed
8.7 minutes
Jul 28, 5:30 PM
Verified
marketing_ads
qualified_pipeline_per_dollar · Primary
hubspot-production
crm.pipeline.read
Every 1h
11 complete · 1 pending · 0 failed
1.42 ratio
Jul 28, 5:00 PM
Verified
marketing_content_creation
qualified_content_assists · Primary
notion-content
content.analytics.read
Every 1h
7 complete · 0 pending · 1 failed
Waiting for first sampleStale

product_activation_recovery · qualified_activation_rate

Evaluated Jul 28, 5:05 PM

Improvedobserved94% confidence
Baseline
51.2
Observed
68.4
Change
+33.6%

Guardrails: 2 of 2 passed.

support_priority_triage · urgent_first_response_minutes

Evaluated Jul 28, 5:35 PM

Target metobserved91% confidence
Baseline
18.5
Observed
8.7
Change
-53.0%

Guardrails: 1 of 1 passed.

marketing_ads · qualified_pipeline_per_dollar

Evaluated Jul 28, 5:05 PM

Improvedmodeled64% confidence
Baseline
1.1
Observed
1.42
Change
+29.1%

Guardrails: 1 of 1 passed.

  • The preview uses modeled attribution until a warehouse source is connected.

marketing_content_creation · qualified_content_assists

Evaluated Jul 28, 5:05 PM

Incompleteincomplete28% confidence
Baseline
Observed
Change

Guardrails: 0 of 1 passed.

  • The latest content analytics sample is stale.
  • No complete comparison window exists.

Connector reconciliation · Degraded

Checked Jul 28, 5:32 PM

Warning

The Notion content connection has not reported health within its freshness window.

Repair: Refresh the Hermes-managed credential and run connection reconciliation again.