Coherence Ecosystem · Production

Unified Reasoning
Engine

Every project generates adoption signals in isolation. Heyday standardises, weights, and reasons over them to produce actionable recommendations — federated across 9 domains.

Access Dashboard → Enroll your domain
📡
Signal
Collection
🔗
Causality
Modelling
🎲
Outcome
Simulation
Scoring &
Gating
🏷
Signed
Recommendation
9
Live Domains
11
Correlation Pairs
147
Test Cases
15
Min Signal Cycle
LIVE
VM 211 · heyday.multiskilled.ai
Every project interprets adoption signals in isolation. Heyday makes cross-project adoption intelligence a shared, composable resource — sovereign and ecosystem-native.

Harvey tracks operator churn. DANAI tracks cascade routing. CE tracks policy drift. MMB tracks brand growth. None of them talk to each other — until now. Heyday proved this model at MMB scale across 50+ brands, then federated the same reasoning engine across the entire Coherence Ecosystem.

Same engine. Different tuning.

The reasoning core is domain-agnostic. Only the signal weights, confidence thresholds, and simulation bias change per domain.

📡

Signal Collection

8 ecosystem adapters pull live metrics every 15 minutes via Promise.allSettled. A failed adapter never blocks others. Signals normalised to 0.0–1.0.

🔗

Causality Modelling

Domain tuning weights applied multiplicatively. The dominant weighted signal drives the recommendation action string. 11 cross-domain correlation pairs detect shared causes.

🎲

Outcome Simulation

Three-scenario Monte Carlo: optimistic (25%), expected (50%), pessimistic (25%). simulation_bias calibrates domain optimism — MMB: 1.2×, CE: 0.9×.

Confidence Gating

Only recommendations above the domain’s confidence_threshold are emitted. Below-threshold signals return null — no action, no noise.

🏷

W3C VC Signing

Every recommendation emitted as a HeydayRecommendationCredential signed with Ed25519. NTFY push fires on confidence ≥ 0.85. FHF propagated to CE on new cross-domain finds.

🔄

Outcome Feedback

Ecosystem peers confirm outcomes via SDK confirmOutcome(). After 5+ confirmed outcomes, Heyday proposes tuning weight adjustments — HITL-reviewed before applying.

Parameterised per project. Identical core.

Domain tuning is the only thing that changes. Add a tuning-{domain}.json and a line in listDomains() — the engine is already there.

harveyLIVE
operator_churn3.0×
churn_alert_density2.5×
retention2.5×
confidence ≥ 0.60 · horizon 14 days · 7 feeds
mmbLIVE
feasibility2.5×
market_size2.0×
brand_growth_rate1.8×
confidence ≥ 0.60 · horizon 90 days · 50+ brands
ceLIVE
governance2.0×
policy_drift2.0×
confidence ≥ 0.75 · horizon 14 days
danaiLIVE
cascade_routing2.0×
learning1.5×
confidence ≥ 0.68 · horizon 7 days
gcuLIVE
portfolio_health2.5×
hitl_backlog2.2×
trust_drift1.8×
confidence ≥ 0.70 · horizon 14 days
sentinelLIVE
posture_score3.0×
anomaly_rate2.5×
policy_drift2.2×
confidence ≥ 0.65 · horizon 3 days
dcdanLIVE
infra_drift2.5×
capability_success_rate2.2×
keeper_graduation2.0×
confidence ≥ 0.60 · horizon 7 days
heydayLIVE
operator_churn3.0×
retention2.0×
confidence ≥ 0.70 · horizon 30 days
socialCONFIGURED
post_success_rate2.5×
content_cadence_health2.0×
engagement_velocity1.8×
confidence ≥ 0.60 · horizon 7 days · Postiz

11 correlation pairs detect shared causes.

When two signals from different domains exceed 0.5 simultaneously, Heyday surfaces a synthesis — a shared root cause invisible to either project alone. Detected correlations are propagated to CE via FHF.

operator_churntier_escalation_rate
Response latency bottleneck
retentionskill_discovery_rate
Onboarding quality is the lever
churn_alert_densitypolicy_drift
Policy friction causing disengagement
anomaly_rateoperator_churn
Environmental instability driving churn
posture_scoreretention
Unsafe environment drives operator attrition
hitl_backlogqueue_age
Human bandwidth is the constraint
post_success_ratebrand_growth_rate
Publishing consistency compounds network effects
policy_driftconflict_rate
Governance instability → operational friction

Built with the ecosystem.

Heyday is a shared resource — designed with the projects it serves, not for them.

MMB
Co-architect · Tuning reference
Opportunity discovery across 50+ brands. The model was proved at MMB scale before it was federated. SCO f8d8b3af resolved.
✓ enrolled · d647ff9
DANAI
Co-chair · Engine architecture
Cascade routing and learning velocity signals. Co-chairs the engine architecture design. Push module live every 30 min.
✓ active · signals live
CE
Governance principal · Engine approval
Governance health and policy drift. CE is the approval authority for all domain enrollments via HITL-D profile 92d289c5.
✓ W1 demands complete
Harvey
Edge provider · Data source
7 live feeds: analytics, operator skills, onboarding cohort, referral stats, FHF, churn alerts, conflict patterns.
✓ 7/7 feeds · push every 15 min

Two paths into the ecosystem.

If your domain is already registered, push signals directly:

import { HeydayClient } from '@coherence/heyday';

const client = new HeydayClient({
  endpoint: 'https://heyday.multiskilled.ai',
  authToken: process.env.HEYDAY_AUTH_TOKEN,
});

// Push signals
await client.ingestSignal('harvey', [{
  key: 'operator_churn', value: 0.72,
  source: 'harvey-desktop',
  collected_at: new Date().toISOString(),
}]);

// Get a recommendation
const { recommendation } = await client.recommend('harvey');
console.log(recommendation.action, recommendation.confidence);

New domain? Submit an enrollment request (HITL-gated, typically 1–2 days):

# Via HTTP
curl -X POST https://heyday.multiskilled.ai/api/v1/enroll \
  -H "Content-Type: application/json" \
  -d '{
    "domain": "my-project",
    "requester_did":
      "did:web:my-project.multiskilled.ai:ecosystem",
    "capabilities": ["my-project.signal.v1"],
    "tuning": {
      "signal_weights": { "adoption_rate": 2.0 },
      "confidence_threshold": 0.65,
      "time_horizon_days": 14
    }
  }'

# Or include heyday: { domain } in peer.announced
# Heyday auto-submits the enrollment for you

Adoption intelligence,
shared across the ecosystem.

The operator dashboard shows live signals, active correlations, recommendation history, and pending enrollments. Authentik-protected — request access from dkm.