SDK v0.5 — Now Available

The AI Assurance Platform

Open-source Compliance-as-Code for High-Risk AI. Transform your Python code into regulatory evidence from the very first line of code.

Install SDKRead the docs
7 Evidence Probes: Trace, Hardware, Carbon, BOM, Integrity, Artifact, Handshake
AI Assurance Audit — Human-readable
Audit Log

Fairness Audit

Scenario: loan_scoring_v2

VERIFIEDe55ae8f096fd
Population Analyzed
690
Male
310
Female
Data Quality
PASS
0.43 > 0.20
80% Rule
PASS
0.82 > 0.80
Stat. Parity
PASS
Δ = 0.13 < 0.20
k-Anonymity
PASS
k = 5 ≥ 3
Equal Opportunity
PASS
Δ = 0.04 < 0.10
🎉

Compliance Assurance: Verified

Policy enforced · 7 probes collected · Evidence vault ready.

Logs your compliance team can review and evidence for EU AI Act Annex IV.

AI Assurance Engine

Import. Enforce. Verify.

Don't guess the rules. Import OSCAL policies. Your CI/CD breaks if the model is non-compliant.

Policy

OSCAL Policy Engine

Import machine-readable policies defined by Legal. If bias exceeds threshold, the pipeline fails. NIST OSCAL standard.

Learn more
Probes

7 Evidence Probes

Trace, Hardware, Carbon, BOM, Integrity, Artifact, Handshake. Automatic evidence collection during training and inference.

Learn more
Metrics

35+ Assurance Metrics

Fairness (binary & multiclass), privacy (k-anonymity, l-diversity), performance, data quality, causal fairness, ESG.

Learn more
Transparency

Glass Box Governance

Educational audit trails. Every metric explains why it matters for regulators. No more black boxes.

Learn more
Privacy

Local Sovereignty

Zero-cloud enforcement. Your data never leaves your machine. Strict mode for CI/CD. Evidence stored locally.

Learn more
Security

Supply Chain & BOM

CycloneDX ML-BOMs, SHA-256 signed artifacts, dependency scanning. Full supply chain visibility for auditors.

Learn more
Start in minutes

Quickstart (Python)

Install with pip or uv, point to your OSCAL policy, and run enforcement in notebooks or CI/CD jobs.

Terminal
$pip install venturalitica

Works with pip / uv / poetry.

1

Install

pip install venturalitica

2

Monitor your Session

with vl.monitor("training"): opens an AI governance session.

3

Enforce Policy

Call vl.enforce to run checks and generate evidence.

Run enforcement
Python
import venturalitica as vl
import pandas as pd

# 1. Load OSCAL Policy (The "Contract")
policy = "risks.oscal.yaml"

# 2. Monitor Training Session
with vl.monitor("loan_scoring_v2"):
    model.fit(X_train, y_train)

# 3. Enforce & Generate Evidence
results = vl.enforce(
    data=df,
    policy=policy,
    target="loan_status",
    gender="Attribute9"
)

Python example; plug your model outputs and policy file.

Output
$python audit_pipeline.py
Venturalítica v0.5.0 :: AI Assurance Engine
----------------------------------------
CONTROL
DESCRIPTION
ACTUAL
RESULT
data-quality-imb...
Data Quality: Minority gr...
0.429
✓ PASS
fairness-80-rule
80% Rule (Gender Fairnes...
0.818
✓ PASS
statistical-parity
Statistical Parity Diff...
0.134
✓ PASS
k-anonymity
k-Anonymity (Privacy Asse...
5
✓ PASS
equal-opportunity
Equal Opportunity Diff...
0.041
✓ PASS
----------------------------------------
✔ Compliance Proof Generated:.venturalitica/runs/latest/results.json
ℹ Evidence Vault:.venturalitica/runs/latest/
View full docs

Contact Us