Skip to content

Industry Wizards

Domain-specific AI assistants with built-in security, compliance, and industry best practices.

Overview

Empathy Framework includes industry-specific wizards that provide:

  • Built-in Security - PII scrubbing, secrets detection, audit logging
  • Domain Knowledge - Industry-specific prompts and workflows
  • Compliance Ready - HIPAA, SOC2, GDPR, industry regulations
  • Easy Integration - Drop-in components for any application

Quick Start

Choose Your Industry

Click the tab for your industry to see the specialized wizard documentation.

Healthcare Wizards

17 HIPAA-compliant AI assistants for medical applications with enhanced PHI protection.

Key Features

  • Enhanced PHI Protection - 10+ medical patterns (MRN, Patient ID, DOB, etc.)
  • Mandatory Encryption - AES-256-GCM for all PHI
  • 90-Day Retention - HIPAA §164.528 compliance
  • Comprehensive Audit Trail - HIPAA §164.312(b) compliant
  • $2M+ Annual ROI - For 100-bed hospitals

Quick Example

from empathy_llm_toolkit import EmpathyLLM
from empathy_llm_toolkit.wizards import HealthcareWizard

# Initialize with security enabled
llm = EmpathyLLM(
    provider="anthropic",
    api_key=os.getenv("ANTHROPIC_API_KEY"),
    enable_security=True
)

# Create HIPAA-compliant wizard
wizard = HealthcareWizard(llm)

# Process patient information (PHI is automatically scrubbed)
result = await wizard.process(
    user_input="Patient John Doe (MRN 123456) needs follow-up for diabetes",
    user_id="doctor@hospital.com"
)

# PHI was removed before sending to LLM
print(result['security_report']['phi_removed'])  # ['mrn', 'name']
What PHI Patterns Are Detected?

Standard PII: - Email addresses - Phone numbers - SSN - Physical addresses - Credit card numbers - IP addresses

Healthcare-Specific PHI: - MRN - Medical Record Numbers - Patient IDs - Patient identifiers - DOB - Dates of birth - Insurance IDs - Insurance/policy numbers - Provider NPI - National Provider Identifiers - CPT Codes - Medical procedure codes - ICD Codes - Diagnosis codes - Medications - Drug names (optional, configurable)

Clinical Handoff (SBAR Protocol) 
wizard = HealthcareWizard(llm)

# Generate SBAR handoff report
result = await wizard.generate_handoff(
    patient_id="PT123456",
    protocol="SBAR",  # Situation, Background, Assessment, Recommendation
    handoff_type="shift_change"
)

print(result['sbar_report'])
# Output:
# **Situation:** 65yo male, chest pain x2h, vitals stable
# **Background:** Hx of MI 2018, on aspirin, metoprolol
# **Assessment:** Possible STEMI, EKG shows ST elevation
# **Recommendation:** Activate cath lab, continue monitoring

HIPAA Compliance Requirements

To maintain HIPAA compliance:

  1. ✅ Enable security: EmpathyLLM(enable_security=True)
  2. ✅ Use encryption at rest for stored data
  3. ✅ Review audit logs daily
  4. ✅ Implement access controls
  5. ✅ Sign Business Associate Agreement with LLM provider

See Also: SBAR Clinical Handoff Example

Finance Wizard

SOC2-compliant AI assistant for financial services with enhanced PII/PCI protection.

Key Features

  • PCI DSS Compliance - Credit card detection and masking
  • Financial PII - Account numbers, routing numbers, SSN
  • Risk Analysis - AML, fraud detection, compliance checks
  • Audit Trail - SOC2 Type II compliant logging

Quick Example

from empathy_llm_toolkit.wizards import FinanceWizard

wizard = FinanceWizard(llm)

# Analyze transaction for compliance
result = await wizard.analyze_transaction(
    transaction_data={
        "amount": 15000,
        "source_account": "****1234",
        "destination_account": "****5678",
        "country": "US"
    },
    check_aml=True,
    check_fraud=True
)

if result['flags']:
    print(f"⚠️  Compliance flags: {result['flags']}")
What Financial PII Is Protected?
  • Credit Card Numbers - Full card number detection and masking
  • Account Numbers - Bank account numbers
  • Routing Numbers - ABA routing numbers
  • SSN - Social Security Numbers
  • ITIN - Individual Taxpayer Identification Numbers
  • EIN - Employer Identification Numbers
  • Investment Account IDs - Brokerage account numbers

Risk Analysis Features

The Finance Wizard includes built-in risk analysis:

  • AML (Anti-Money Laundering) - Flags suspicious transactions
  • Fraud Detection - Pattern-based fraud indicators
  • Sanctions Screening - OFAC compliance checks
  • KYC Validation - Know Your Customer verification

AI assistant for legal practices with document classification and privilege protection.

Key Features

  • Attorney-Client Privilege - Automatic privilege detection
  • Document Classification - Contract, brief, discovery types
  • Legal Citation - Find relevant case law
  • Confidentiality - Work product protection

Quick Example

from empathy_llm_toolkit.wizards import LegalWizard

wizard = LegalWizard(llm)

# Analyze legal document
result = await wizard.analyze_document(
    document_text="...",
    document_type="contract",
    jurisdiction="CA"
)

print(result['risk_factors'])
print(result['suggested_clauses'])
Contract Review 
# Review contract for risks
result = await wizard.review_contract(
    contract_text="...",
    contract_type="employment",
    jurisdiction="CA",
    check_for=[
        "non_compete",
        "indemnification",
        "termination",
        "ip_assignment"
    ]
)

# Get risk assessment
for risk in result['risks']:
    print(f"{risk['severity']}: {risk['description']}")
    print(f"Suggested fix: {risk['remediation']}")

Retail Wizard

AI assistant for e-commerce and retail operations.

Key Features

  • Inventory Management - Stock optimization suggestions
  • Pricing Strategy - Dynamic pricing recommendations
  • Customer Service - Support automation
  • Sales Analytics - Trend analysis

Quick Example

from empathy_llm_toolkit.wizards import RetailWizard

wizard = RetailWizard(llm)

# Optimize inventory
result = await wizard.optimize_inventory(
    product_data={
        "sku": "PROD123",
        "current_stock": 50,
        "sales_last_30d": 120,
        "season": "winter"
    }
)

print(result['reorder_quantity'])
print(result['optimal_price'])

Education Wizard

FERPA-compliant AI assistant for educational institutions.

Key Features

  • Student Privacy - FERPA compliance (20 U.S.C. § 1232g)
  • :material-account-student: Student PII Protection - Student IDs, grades, records
  • Assignment Grading - Automated assessment assistance
  • Curriculum Support - Lesson plan generation

Quick Example

from empathy_llm_toolkit.wizards import EducationWizard

wizard = EducationWizard(llm)

# Generate lesson plan (no student PII exposed)
result = await wizard.generate_lesson_plan(
    subject="Mathematics",
    grade_level=8,
    topic="Linear Equations",
    duration_minutes=45
)

print(result['lesson_plan'])
print(result['assessment_questions'])

HR Wizard

AI assistant for human resources with employee PII protection.

Key Features

  • Employee PII Protection - SSN, DOB, salary, benefits
  • Job Descriptions - Generate JD from requirements
  • Resume Screening - Bias-free candidate evaluation
  • Compliance - EEOC, ADA, FLSA guidance

Quick Example

from empathy_llm_toolkit.wizards import HRWizard

wizard = HRWizard(llm)

# Generate job description
result = await wizard.generate_job_description(
    title="Senior Software Engineer",
    department="Engineering",
    level="Senior",
    requirements=["Python", "AWS", "5+ years experience"]
)

print(result['job_description'])

Technology Wizard

AI assistant for software development and IT operations.

Key Features

  • Bug Analysis - Root cause identification
  • Code Review - Security and quality checks
  • Cloud Architecture - AWS/Azure/GCP design patterns
  • Security Scanning - Vulnerability detection

Quick Example

from empathy_llm_toolkit.wizards import TechnologyWizard

wizard = TechnologyWizard(llm)

# Analyze code for security issues
result = await wizard.review_code(
    code=code_snippet,
    language="python",
    check_for=["sql_injection", "xss", "secrets"]
)

for issue in result['security_issues']:
    print(f"{issue['severity']}: {issue['description']}")

Additional Wizards

Accounting Wizard

AI assistant for accounting and bookkeeping - GAAP/IFRS compliance - Financial statement analysis - Tax preparation assistance

Customer Support Wizard

AI assistant for customer service operations - Ticket classification - Response templates - Sentiment analysis

Government Wizard

AI assistant for government agencies - FOIA compliance - Public records management - Citizen service automation

Insurance Wizard

AI assistant for insurance operations - Claims processing - Underwriting assistance - Risk assessment

Logistics Wizard

AI assistant for supply chain and logistics - Route optimization - Inventory forecasting - Shipment tracking

Manufacturing Wizard

AI assistant for manufacturing operations - Production scheduling - Quality control - Equipment maintenance

Real Estate Wizard

AI assistant for real estate professionals - Property valuation - Lease generation - Market analysis

Research Wizard

AI assistant for academic and scientific research - Literature review - Citation management - Data analysis

Sales Wizard

AI assistant for sales teams - Lead qualification - Proposal generation - CRM integration

Base Wizard API

All wizards extend the BaseWizard class with common functionality:

Base class for all Empathy LLM wizards

Provides: - Integration with EmpathyLLM - Security pipeline configuration - Domain-specific prompting - Audit logging - Session management

__init__(llm, config)

Initialize wizard with LLM and configuration

Parameters:

Name Type Description Default
llm EmpathyLLM

EmpathyLLM instance (with or without security enabled)

 required
config WizardConfig

Wizard configuration

 required

get_config()

Get wizard configuration

get_name()

Get wizard name

process(user_input, user_id, empathy_level=None, session_context=None) async

Process user input through the wizard

Parameters:

Name Type Description Default
user_input str

User's message or request

 required
user_id str

Identifier for the user

 required
empathy_level int | None

Override default empathy level (optional)

 None
session_context dict[str, Any] | None

Additional context for the conversation

 None

Returns:

Type Description
dict[str, Any]

Dictionary containing: - response: AI response - empathy_level: Level used - security_report: Security scan results (if enabled) - metadata: Additional wizard metadata

WizardConfig

Configuration for an Empathy wizard

Source code in empathy_llm_toolkit/wizards/base_wizard.py 
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
@dataclass
class WizardConfig:
    """Configuration for an Empathy wizard"""

    # Wizard identity
    name: str
    description: str
    domain: str  # healthcare, finance, legal, general, etc.

    # Empathy level (0-4)
    default_empathy_level: int = 2

    # Security configuration
    enable_security: bool = False
    pii_patterns: list[str] = field(default_factory=list)
    enable_secrets_detection: bool = False
    block_on_secrets: bool = True

    # Audit configuration
    audit_all_access: bool = False
    retention_days: int = 180

    # Classification
    default_classification: str = "INTERNAL"  # PUBLIC, INTERNAL, SENSITIVE
    auto_classify: bool = True

    # Memory configuration
    enable_memory: bool = False
    memory_config: ClaudeMemoryConfig | None = None

Configuration options:

  • name (str): Wizard identifier
  • domain (str): Industry domain (healthcare, finance, legal, etc.)
  • default_empathy_level (int): Empathy level 0-4 (default: 2)
  • enable_security (bool): Enable PII/secrets detection
  • pii_patterns (list): Custom PII patterns to detect
  • enable_secrets_detection (bool): Scan for API keys, passwords
  • audit_all_access (bool): Log all wizard interactions
  • retention_days (int): Audit log retention (default: 180 days)
  • default_classification (str): Data classification (PUBLIC, INTERNAL, SENSITIVE)

Creating Custom Wizards

Build Your Own Domain-Specific Wizard

You can create custom wizards for your specific industry:

from empathy_llm_toolkit.wizards import BaseWizard, WizardConfig
from empathy_llm_toolkit import EmpathyLLM

class MyIndustryWizard(BaseWizard):
    """Custom wizard for my industry"""

    def __init__(self, llm: EmpathyLLM):
        config = WizardConfig(
            name="my_industry",
            domain="custom",
            description="AI assistant for my industry",
            enable_security=True,
            pii_patterns=["custom_pattern"],
            default_classification="INTERNAL"
        )
        super().__init__(llm, config)

    async def process(self, user_input: str, user_id: str):
        """Custom processing logic"""

        # Add domain-specific prompts
        enhanced_prompt = f"""
        You are an AI assistant specialized in {self.config.domain}.

        User request: {user_input}
        """

        # Use parent LLM with security enabled
        response = await self.llm.interact(
            user_id=user_id,
            prompt=enhanced_prompt,
            context={"wizard": self.config.name}
        )

        return response

# Use your custom wizard
llm = EmpathyLLM(provider="anthropic", api_key="...")
wizard = MyIndustryWizard(llm)

result = await wizard.process(
    user_input="Help me with industry-specific task",
    user_id="user@company.com"
)

Security Best Practices

Production Security Checklist

For all wizards in production:

  • [ ] Enable security features: enable_security=True
  • [ ] Configure appropriate PII patterns for your industry
  • [ ] Enable secrets detection: enable_secrets_detection=True
  • [ ] Enable audit logging: audit_all_access=True
  • [ ] Set correct data classification
  • [ ] Review audit logs regularly
  • [ ] Test PII scrubbing before production
  • [ ] Implement access controls
  • [ ] Encrypt data at rest
  • [ ] Sign appropriate compliance agreements (BAA for HIPAA, DPA for GDPR)

Classification Levels

PUBLIC - No PII, can be shared publicly

INTERNAL - Internal business data, PII scrubbed

SENSITIVE - PHI, financial data, legal privileged - requires encryption

See Also