Developers | Unisapience.ai

Python SDK

pip install unisapience

TypeScript SDK

npm install @unisapience/sdk

REST + Webhooks

api.unisapience.ai/v1

Terraform Provider

registry.terraform.io

Authentication

All API requests are authenticated with a Bearer token passed in the Authorization header. Keys are scoped by tier and rotate on a 90-day schedule by default.

No cookies, no sessions — stateless token auth

TLS 1.3 required on all endpoints

Automatic key rotation with 7-day overlap window

Scoped keys: read-only, write, admin

curl example bash

# Compute extended composite SSS (SSPLX-002)
curl https://api.unisapience.ai/v2/composite/extended \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model_id": "acme-underwriting-v3",
    "capability_metric": { "type": "compute_flops", "value": 1.2e25 },
    "meta_conditions": {
      "KILL": true, "SAF": true, "HITL": true,
      "AUT": true, "TRU": true, "MAN": true
    },
    "registers": ["statics","mechanics","dynamics","thermodynamics","kinematics"]
  }'

API Reference · v2

Endpoints

Base URL: https://api.unisapience.ai/v2

Legacy seven-factor SSS endpoints remain available under /v1. Extended SSPLX-002 endpoints under /v2 are the recommended path for new integrations.

POST /composite/extended Compute the extended composite SSS with full SSPLX-002 decomposition

Request body

{
  "model_id": "acme-underwriting-v3",
  "capability_metric": {
    "type": "compute_flops",
    "value": 1.2e25,
    "reference": 1.0e24
  },
  "meta_conditions": {
    "KILL": true, "SAF": true,
    "HITL": true, "AUT": true,
    "TRU": true, "MAN": true
  },
  "registers": ["statics","mechanics","dynamics",
                "thermodynamics","kinematics"],
  "normalizer_form": "exponential_decay",
  "shadow_distance_mode": "geometric"
}

Response 200

{
  "model_id": "acme-underwriting-v3",
  "sss_extended": 235,
  "coherence_score": 765,
  "band": "drift",
  "capability_normalizer": 0.85,
  "scale": 1000,
  "registers": {
    "statics": 180, "mechanics": 220,
    "dynamics": 280, "thermodynamics": 210,
    "kinematics": 240
  },
  "meta_vetoed": false,
  "meta_status": {
    "KILL": "pass", "SAF": "pass",
    "HITL": "pass", "AUT": "pass",
    "TRU": "pass", "MAN": "pass"
  },
  "alignment_edge_excluded": true,
  "framework_version": "sspx-002"
}

POST /capability/normalize Compute C(κ) for a given capability metric and functional form

Parameters

kappa

number*Capability metric value

kappa_ref

numberReference capability level

form

enumlinear_inverse · exponential_decay · threshold_tiered · velocity_ratio · hybrid

alpha

numberDecay coefficient (exponential form)

tier_table

objectTier→multiplier map (threshold form)

Response 200

{
  "capability_normalizer": 0.85,
  "kappa": 1.2e25,
  "kappa_ref": 1.0e24,
  "form": "exponential_decay",
  "alpha": 0.5,
  "interpretation":
    "Safety progress tracking capability;
     mild penalty applied.",
  "alert": null
}

GET /registers/score Five-register sub-score vector for a registered model

Parameters

model_id

string*Registered model identifier

registers

arraySubset of statics, mechanics, dynamics, thermodynamics, kinematics

window

stringTrajectory window (7d, 30d, 90d)

Each register sub-score is computed under a register-specific weight profile reflecting failure-mode characteristic of that time horizon.

Response 200

{
  "model_id": "acme-underwriting-v3",
  "registers": {
    "statics":        { "score": 180, "horizon": "t=0",          "dominant": "alignment" },
    "mechanics":      { "score": 220, "horizon": "input",         "dominant": "robustness" },
    "dynamics":       { "score": 280, "horizon": "trajectory",    "dominant": "alignment+control" },
    "thermodynamics": { "score": 210, "horizon": "irreversible",  "dominant": "control" },
    "kinematics":     { "score": 240, "horizon": "velocity",      "dominant": "control" }
  },
  "trajectory_alert": null,
  "divergence_pattern": "stable"
}

GET /meta/check Evaluate transcendental meta-condition veto layer (∞)

Meta-conditions

KILLKill-switch available within time bound

SAFSafety boundary defined

HITLHuman-in-loop on critical decisions

AUTAutomation-bias mitigation

TRUTrust calibration posture

MANManipulation resistance

Any meta-condition returning false collapses the composite SSS to its veto value regardless of all other measurements. Veto events emit a categorical alert independent of graduated scoring.

Response 200

{
  "model_id": "acme-underwriting-v3",
  "vetoed": false,
  "indicators": {
    "KILL": { "status": "pass", "evidence": "kill_switch_audit_v3" },
    "SAF":  { "status": "pass", "evidence": "safety_boundary_spec_2026q2" },
    "HITL": { "status": "pass", "evidence": "hitl_coverage_98pct" },
    "AUT":  { "status": "pass", "evidence": "automation_bias_controls_v2" },
    "TRU":  { "status": "pass", "evidence": "trust_calibration_audit_2026q1" },
    "MAN":  { "status": "pass", "evidence": "manipulation_resist_pen_test" }
  },
  "categorical_alerts": []
}

GET /matrix/score Per-primitive scores across the 9 × 6 dyadic-group × compositional-period matrix

Parameters

model_id

string*Target model

groups

arraySubset of I–IX (default: all)

periods

arraySubset of P1–P6 (default: all)

compounds

boolInclude derived compound primitives (ERR, STR, VAL, RED, ROL, TLS, DEL, GOAL, REV, MISU)

Groups I–IX enumerate the nine scored pairwise edges of the rectified pentachoron substrate. The A–Ω alignment edge is structurally excluded as irreducible. Periods P1–P6: Primordial, Relational, Synthetic, Mediated, Social, Absolute.

Response 200 · excerpt

{
  "model_id": "acme-underwriting-v3",
  "matrix": {
    "I.P1":  { "primitive": "HAL", "score": 0.18, "shadow_d": 1.4 },
    "I.P2":  { "primitive": "BIA", "score": 0.22, "shadow_d": 1.2 },
    "II.P3": { "primitive": "DEP", "score": 0.31, "shadow_d": 0.9 },
    "III.P2":{ "primitive": "JBR", "score": 0.27, "shadow_d": 1.1 },
    "IV.P1": { "primitive": "POL", "score": 0.19, "shadow_d": 1.3 },
    "IV.P6": { "primitive": "ETH", "score": 0.34, "shadow_d": 0.8 }
    /* ... 48 more primitives ... */
  },
  "group_subscores": {
    "I": 215, "II": 245, "III": 268, "IV": 290, "V": 220,
    "VI": 240, "VII": 195, "VIII": 280, "IX": 260
  },
  "period_subscores": {
    "P1": 200, "P2": 215, "P3": 245,
    "P4": 240, "P5": 270, "P6": 295
  },
  "compounds_active": ["MISU", "DEL"],
  "alignment_edge_excluded": true
}

GET /trajectory/alerts SSS, register, and capability-normalizer deltas over rolling window

Parameters

model_id

string*Target model

window

string7d, 30d, 90d (default: 30d)

threshold

numberSSS delta to trigger (default: 50)

Response 200

{
  "model_id": "acme-underwriting-v3",
  "window": "30d",
  "sss_start": 215,
  "sss_current": 235,
  "sss_delta": 20,
  "kappa_delta": 0.15,
  "ds_dt": 0.04,
  "dk_dt": 0.18,
  "velocity_ratio": 0.22,
  "alert_level": "watch",
  "alerts": [
    {
      "type": "capability_outpacing_safety",
      "register": "dynamics",
      "detected_at": "2026-05-22T08:14:00Z",
      "description":
        "Capability gain rate exceeds safety gain rate; C(κ) trend declining."
    }
  ]
}

GET /compliance/gap-check Regulatory gap analysis mapped to SSPLX axes

Parameters

model_id

string*Target model

frameworks

arrayeu_ai_act, nist_rmf, iso_42001, sec, hipaa, dora, sr_11_7

Response 200

{
  "model_id": "acme-underwriting-v3",
  "gaps": [
    {
      "framework": "eu_ai_act",
      "article": "9",
      "requirement": "Risk management system",
      "sspx_axis": "register.dynamics",
      "status": "partial",
      "severity": "high",
      "remediation":
        "Implement continuous risk monitoring per Art. 9(4)."
    }
  ],
  "compliant_count": 14,
  "gap_count": 3
}

POST /models/register Register a model for ongoing SSPLX-002 monitoring

Request body

{
  "name": "Acme Underwriting v3",
  "model_type": "llm_fine_tuned",
  "base_model": "gpt-4o",
  "deployment_env": "production",
  "regulatory_context": "financial_services",
  "owner_team": "risk-ai@acme.com",
  "capability_baseline": {
    "type": "compute_flops",
    "value": 1.0e24
  },
  "aibom": { ... }
}

Response 201

{
  "model_id": "acme-underwriting-v3",
  "registered_at": "2026-05-14T09:00:00Z",
  "monitoring_status": "active",
  "framework_version": "sspx-002",
  "initial_sss": null,
  "first_score_eta": "2026-05-14T11:00:00Z",
  "webhook_configured": false
}

Code samples

Integrate in minutes

Python

pip install unisapience

from unisapience import Client, CapabilityMetric, MetaConditions

# Initialize with your API key
client = Client(api_key="sk_live_...")

# Compute the extended composite
result = client.composite.extended(
    model_id="acme-underwriting-v3",
    capability_metric=CapabilityMetric(
        type="compute_flops",
        value=1.2e25,
        reference=1.0e24,
    ),
    meta_conditions=MetaConditions(
        KILL=True, SAF=True,
        HITL=True, AUT=True,
        TRU=True, MAN=True,
    ),
    normalizer_form="exponential_decay",
)

print(f"SSS: {result.sss_extended}  CS: {result.coherence_score}")
print(f"Band: {result.band}  C(κ): {result.capability_normalizer}")

# Register vector
for reg, score in result.registers.items():
    print(f"  {reg}: {score}")

# Meta-condition veto check
if result.meta_vetoed:
    print("⚠ Composite vetoed — see meta_status for failing prerequisite.")

TypeScript

npm i @unisapience/sdk

import { UnisapienceClient } from '@unisapience/sdk';

const client = new UnisapienceClient({
  apiKey: process.env.UNISAPIENCE_API_KEY,
});

// Compute the extended composite
const result = await client.composite.extended({
  modelId: 'acme-underwriting-v3',
  capabilityMetric: { type: 'compute_flops', value: 1.2e25 },
  metaConditions: {
    KILL: true, SAF: true, HITL: true,
    AUT: true, TRU: true, MAN: true,
  },
  registers: ['statics', 'mechanics', 'dynamics',
              'thermodynamics', 'kinematics'],
});

console.log(`SSS: ${result.sssExtended} (CS ${result.coherenceScore})`);
console.log(`Band: ${result.band}  C(κ): ${result.capabilityNormalizer}`);

// Subscribe to capability-outpacing-safety events
await client.webhooks.create({
  modelId: 'acme-underwriting-v3',
  url: 'https://your-app.com/api/sss-alerts',
  events: [
    'composite.band_change',
    'meta.veto',
    'capability.outpacing_safety',
    'register.divergence',
  ],
});

Webhooks

Subscribe to real-time events and push SSS signals into your GRC platform, SIEM, or alerting pipeline. All payloads are signed with HMAC-SHA256.

meta.veto Categorical alert — a meta-condition transitioned to breached; composite collapsed.

capability.outpacing_safety dC/dκ exceeds safety-progress velocity; C(κ) trend declining.

composite.band_change Model crosses a band boundary (Drift → Mixed Equilibrium, etc.)

register.divergence Register vector trajectory shows diagnostically significant divergence pattern.

compliance.gap Regulatory gap status changes for a monitored model.

score.weekly Weekly composite + register + meta delta report for all registered models.

Webhook payload · meta.veto

{
  "event": "meta.veto",
  "timestamp": "2026-05-14T09:45:00Z",
  "model_id": "acme-underwriting-v3",
  "vetoed_condition": "KILL",
  "previous_status": "pass",
  "current_status": "fail",
  "evidence": "kill_switch_audit_failed_2026-05-14",
  "composite_before": 235,
  "composite_after": 0,
  "band_before": "drift",
  "band_after": "vetoed",
  "recommended_action":
    "Restore kill-switch within configured time bound; recompute composite.",
  "signature": "sha256=7b3f9c..."
}

Pricing

API tiers

Free

For evaluation & prototyping

$0/mo

100 score requests/month
Up to 2 registered models
Extended composite (read-only)
3 of 5 registers
No capability normalizer
No webhooks

Growth

For production integrations

$499/mo

billed monthly, cancel anytime

5,000 score requests/month
Up to 25 registered models
Full 5-register stratification
Capability normalizer C(κ)
Meta-condition assessment
Webhooks + trajectory alerts
Compliance gap-check

Enterprise

For regulated industries

Custom

Unlimited requests
Unlimited model registry
Full 9×6 matrix scoring
Compound primitive derivation
Custom register profiles
SIEM integration support
SLA + dedicated support
On-prem deployment option

Contact sales

Start integrating

Free tier available with no credit card required. Enterprise early access available — email us for a key.

Request enterprise access

SSPLX-002 risk intelligence, API-first.

Authentication

Endpoints

Integrate in minutes

Webhooks

API tiers

Start integrating