Session-level state-trait decomposition
Every response splits stable trait, slow drift, and state noise per construct. LST-R / STARTS / TSO with a Bayesian-partial cold-start for low-session users — surfaced via the same response.
Neumatics Psychometric API · v4.1
Longitudinal psychometric inference, with reliability metadata on every score.
Versioned REST/gRPC. Trait inference with state-trait decomposition, per-user reliability (SEM · ICC · test-retest), conformal 95% CIs, scalar-invariance metadata, and per-revision calibration provenance. The same engine that powers SoulMap.app.
Reliability metadata on every score
SEM, ICC(2,1), 2-week and 8-week test-retest, n_sessions_used, IRT model. Conformal 95% CIs with empirical coverage held in [94%, 96%].
Measurement invariance as a release gate
Configural / metric / scalar invariance across device, language, age cohort, and (with opt-in) neurodivergence. CI-enforced on every factory revision.
Closed dataset · publishable methodology
Every revision has a published diff and a calibration trail. Customers can subscribe to the revision feed; raw data never leaves the Neumatics ecosystem.
A response, not a score.
Every Tier A construct ships with reliability, decomposition, invariance, and evidence-source attribution. No competitor exposes this depth.
{
"user_id": "opaque-hash",
"scored_at": "2026-12-01T14:22:00Z",
"constructs": {
"big5_neuroticism": {
"tier": "a",
"presented_as": "Emotional Sensitivity",
"score": 0.68,
"scale": "z_score",
"ci_95": [0.54, 0.82],
"ci_method": "conformal_split",
"reliability": {
"sem": 0.07,
"icc_2_1": 0.82,
"test_retest_r_8wk": 0.79,
"n_sessions_used": 11
},
"state_trait_decomposition": {
"method": "lst_r",
"stable_component": 0.61,
"state_variance_absorbed": 0.12
},
"invariance": {
"device": "scalar",
"language": "metric",
"cohort_age": "configural"
},
"instrument_citation": "BFI-2 / IPIP-NEO-120"
}
},
"model_version": "nexus-v4.1.2",
"rule_set_version": "nexus-rules-v4.1.0",
"vocabulary_version": "nexus-vocab-v1.2.0"
}Trust posture
- No-train guarantee — your data never trains shared models.
- Per-subject deletion (GDPR DSAR) cascades Firestore + BigQuery + traces.
- EU region pinning at Enterprise (europe-west12 / de-central1).
- SOC 2 Type II audit on roadmap; HIPAA BAA available at Enterprise.
- VPC-SC perimeter; CMEK at rest; A2A agent card published.
- Quarterly methodology releases with human sign-off.
Limited-risk under EU AI Act. Hiring and clinical use cases gated. 18+ consumer surface; persona factory excludes minors from trait-scoring distributions.
Quickstart
The API is in private alpha as of staging. The shape below is the stable v1 contract.
curl https://api.nexus.soulmap.app/v1/users/u_abc/score \
-H "Authorization: Bearer $NEXUS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"constructs": ["big5", "attachment", "emotion_granularity"]}'