guide-framework 1.1.0

GUIDE Framework – Ethical AI Governance, Explainability (XAI), Transparency & Responsible AI Auditing.

GUIDE Framework – Ethical AI, Governance & Explainability (XAI)

guide-framework is a modern Ethical AI toolkit that helps organizations audit, explain, and govern AI systems using the G.U.I.D.E. ethical principles:

🌟 GUIDE AI ETHICS FRAMEWORK

G – Governance
U – Universal Design
I – Identification
D – Dignity
E – Equity

This framework ensures:

• Fairness
• Transparency
• Explainability
• Accessibility (508/WCAG)
• Responsible AI decision-making
• Ethical auditing and governance

It includes both Ethical AI auditing and XAI-based explainability tools, plus a built-in demo audit.

---

🚀 INSTALLATION

pip install guide-framework

Requires Python 3.8+

For data-dependent audits:

pip install pandas scikit-learn numpy

---

🧠 QUICK START — ETHICAL AI AUDIT

from guide_framework import EthicalAIAudit

audit = EthicalAIAudit("LoanModel")

audit.transparency_check({
    "model_name": "LoanModel",
    "version": "1.0",
    "training_data": "internal",
    "documentation": True
})

audit.fairness_check("Evaluate this loan application.")
audit.safety_check("This system may present danger.")
audit.accessibility_check("This explanation is readable and clear.")

print(audit.finalize_report())

Output example:

{
  "model": "LoanModel",
  "results": {
    "transparency": {"status": "pass"},
    "fairness": {"status": "pass"},
    "safety": {"status": "fail"},
    "accessibility": {"status": "pass"}
  },
  "score": "3/4",
  "approved": true
}

---

🎧 EXPLAINABILITY (XAI)

from guide_framework import ExplainabilityGuide

xai = ExplainabilityGuide(api_key="YOUR_OPENAI_KEY")

explanation = xai.explain_decision(
    "The system rejected the loan.",
    context="Applicant missing income documentation"
)

print(explanation)

---

🧾 TRANSPARENCY SUMMARY

metadata = {
    "model_name": "LoanModel",
    "purpose": "Predict risk",
    "training_data": "public datasets",
    "version": "1.0.0",
    "risks": "May underperform for seniors"
}

print(xai.transparency_summary(metadata))

---

🧪 TEST AUDIT SCRIPT (test_audit.py)

Run a full ethical audit in one file.

📄 Create test_audit.py

from guide_framework import EthicalAIAudit

# Create an audit instance
audit = EthicalAIAudit("CreditCardAI")

# Run tests
audit.transparency_check({
    "model_name": "CreditCardAI",
    "version": "2.1",
    "training_data": "bank transactions 2018–2024",
    "documentation": True
})

audit.fairness_check("Applicant includes gender and ethnicity attributes")
audit.safety_check("No dangerous content here")
audit.accessibility_check("This message is readable.")

# Generate final structured output
report = audit.finalize_report()

print("\n=== FINAL AUDIT REPORT ===")
print(report)

▶️ Run it:

python test_audit.py

---

▶️ BUILT-IN TEST AUDIT DEMO

A demo audit is included inside the framework.

Run it directly:

python -m guide_framework.run_demo

You’ll see:

=== GUIDE FRAMEWORK DEMO AUDIT ===
{ ... audit results ... }

Perfect for training, compliance reviews, and onboarding.

---

📦 MODULE MAP

guide_framework/
  ├─ auditor.py           # Ethical audit (transparency, safety, fairness, 508)
  ├─ xai_meaning.py       # XAI explainability & transparency
  ├─ governance.py        # G – Governance
  ├─ understanding.py     # U – Universal Design
  ├─ integrity.py         # I – Identification / Integrity
  ├─ disclosure.py        # D – Disclosure
  ├─ equity.py            # E – Equity
  ├─ situation_templates.py
  ├─ knowledge_base.py
  ├─ guide_simple.py
  ├─ guide.py
  ├─ run_demo.py          # ★ Built-in demo audit script
  └─ other supporting utilities…

---

📋 GUIDE ETHICAL CHECKLIST

G – Governance

☑ Oversight structure
☑ Documentation
☑ Model lifecycle control

U – Universal Design

☑ Inclusive testing
☑ Diverse demographic validation

I – Identification

☑ Clear indication of AI-generated content

D – Dignity

☑ Respectful outputs
☑ Human-centered safety

E – Equity

☑ Fairness across groups
☑ Bias detection

---

🧩 FAIRNESS RECOMMENDATIONS

1. Remove sensitive demographic fields
2. Review group disparities quarterly
3. Add human review to borderline cases
4. Provide explanations to affected users
5. Test performance across culture, age, gender
6. Establish ethical escalation procedures

---

🔐 API KEYS & SECURITY

Your OpenAI key is not stored.

xai = ExplainabilityGuide(api_key="sk-xxxx")

or:

export OPENAI_API_KEY="sk-xxxx"

---

⚠ TROUBLESHOOTING

Upgrade:

pip install --upgrade guide-framework

Rebuild manually:

Remove-Item -Recurse -Force dist, build, *.egg-info
python setup.py sdist bdist_wheel

---

📜 LICENSE

MIT License

---

👤 AUTHOR

Kamal Master
Email: [e-mail]

---

✔ GUIDE AUDIT COMPLETE

Build AI that earns trust through ethical excellence.