shipright 0.1.0

Open-source AI model monitoring with automated compliance documentation

ShipRight

Open-source AI model monitoring with automated compliance documentation.

---

ShipRight monitors your AI models in production — traditional ML and LLMs — and automatically generates audit-ready compliance documentation for NIST AI RMF, the EU AI Act, Colorado CAIA, and other regulations.

3 lines of code. Zero config. Continuous compliance.

import shipright as sr

sr.init(api_key="sr-...", project="fraud-detector")
model = sr.wrap(your_model, reference_data=X_train)

# That's it. Every prediction is now monitored.
predictions = model.predict(X_new)

Why ShipRight?

AI regulations are here. Colorado's AI Act takes effect June 30, 2026 with fines up to $20,000 per violation per consumer. The EU AI Act reaches full application August 2, 2026. Texas, California, and Illinois have active AI laws already.

Most compliance tools cost $50,000+/year and require months to implement. ShipRight gives you:

• Drift detection on autopilot — PSI, KS test, Jensen-Shannon divergence running continuously on your production data
• NIST AI RMF alignment — maps your monitoring to Govern/Map/Measure/Manage functions, providing safe harbor under Colorado, Texas, and California laws
• Auto-generated compliance docs — model cards, impact assessments, risk assessments, and audit trail exports in PDF/DOCX
• LLM monitoring — semantic drift, hallucination detection, PII scanning, and provider change tracking for OpenAI, Anthropic, and open-source models
• < 5ms overhead — async telemetry that won't slow down your inference pipeline

Quick Start

pip install shipright

Traditional ML

import shipright as sr
from sklearn.ensemble import RandomForestClassifier

# Initialize
sr.init(api_key="sr-...", project="fraud-detector")

# Wrap your model — works with sklearn, PyTorch, XGBoost, LightGBM, TensorFlow
model = RandomForestClassifier()
model.fit(X_train, y_train)
monitored = sr.wrap(model, reference_data=X_train)

# Use exactly like before — monitoring is automatic
predictions = monitored.predict(X_new)
probabilities = monitored.predict_proba(X_new)

LLM / GenAI

import shipright as sr

# Wrap OpenAI — one line, zero code changes to your LLM calls
client = sr.wrap_openai(
    api_key="sk-...",
    project="customer-chatbot",
    compliance=["nist-ai-rmf", "eu-ai-act", "colorado-caia"]
)

# Use exactly like the OpenAI client
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Analyze my portfolio"}]
)

RAG Pipelines

import shipright as sr

# Wrap LangChain RAG chains
monitored_chain = sr.wrap_langchain(
    chain=retrieval_qa_chain,
    project="hr-policy-bot",
    compliance=["nist-ai-rmf", "colorado-caia"]
)

result = monitored_chain.invoke({"query": "What is our parental leave policy?"})

What Gets Monitored

Signal	Traditional ML	LLM / GenAI
Feature drift (PSI, KS, JS)	✅	—
Prediction distribution shift	✅	—
Semantic drift (embedding distance)	—	✅
Hallucination / faithfulness	—	✅
PII leakage detection	—	✅
Output consistency	—	✅
Provider drift (silent model updates)	—	✅
Bias / disparate impact	✅	✅
Performance metrics (accuracy, F1, AUC)	✅	—
Latency + cost tracking	✅	✅

Compliance Frameworks Supported

Framework	Status	Safe Harbor
NIST AI RMF (Govern/Map/Measure/Manage)	✅ Supported	CO, TX, CA
EU AI Act (Art. 9–15, Annex III)	✅ Supported	—
Colorado CAIA (SB 24-205)	✅ Supported	✅ NIST
Texas TRAIGA (HB 149)	✅ Supported	✅ NIST
California SB 942 / SB 53 / AB 2013	🔜 Coming	✅ NIST
CCPA ADMT	🔜 Coming	—
ISO 42001	🔜 Coming	—

ShipRight Cloud

The open-source SDK gives you drift detection and alerting for free. ShipRight Cloud adds:

• Auto-generated compliance reports — model cards, CAIA impact assessments, NIST alignment evidence, audit exports
• Safe harbor documentation — proves NIST AI RMF alignment for legal protection
• Consumer disclosure templates — pre-written notice language per jurisdiction
• Multi-model dashboard — compliance status across all your models
• Team collaboration — RBAC, shared dashboards, compliance officer views
• 3-year evidence retention — meets CAIA retention requirements

→ Start free at shipright.run

Configuration

sr.init(
    api_key="sr-...",              # or set SHIPRIGHT_API_KEY env var
    project="my-project",          # project name
    environment="production",      # production | staging | dev
    role="deployer",               # deployer | developer (per CAIA)
    compliance=[                   # frameworks to map against
        "nist-ai-rmf",
        "eu-ai-act",
        "colorado-caia",
    ],
)

Or use environment variables:

export SHIPRIGHT_API_KEY=sr-...
export SHIPRIGHT_PROJECT=my-project
export SHIPRIGHT_ENVIRONMENT=production

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

• 🐛 Report a bug
• 💡 Request a feature
• 💬 Join the discussion

License

Apache 2.0 — see LICENSE for details.

---

Website · Docs · Community · Twitter