Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
This project converts natural language audience intent into Marketing Cloud SQL queries, enabling marketers and developers to build targeted audience segments without deep SQL knowledge. It translates plain English descriptions of target audiences into valid Salesforce Marketing Cloud (SFMC) SQL queries for data extensions.
Installation
git clone https://github.com/alphaparkinc/genpark-marketing-cloud-sql-audience-generator-skill.git
cd genpark-marketing-cloud-sql-audience-generator-skill
pip install -r requirements.txt
Core Functionality
The skill generates SQL queries for Marketing Cloud data extensions based on natural language input. It understands common marketing segmentation criteria like demographics, behavior, engagement, and purchase history.
Basic Usage
from genpark_audience_generator import AudienceGenerator
# Initialize the generator
generator = AudienceGenerator(api_key=os.getenv())
intent =
sql_query = generator.generate_sql(intent)
(sql_query)
'GENPARK_API_KEY'
# Generate SQL from natural language
"customers who purchased in the last 30 days and opened at least 3 emails"
print
Expected output:
SELECT
c.SubscriberKey,
c.EmailAddress,
c.FirstName,
c.LastName
FROM Customers c
INNERJOIN Purchases p ON c.SubscriberKey = p.SubscriberKey
INNERJOIN EmailEngagement e ON c.SubscriberKey = e.SubscriberKey
WHERE p.PurchaseDate >= DATEADD(day, -30, GETDATE())
AND e.EventType ='Open'GROUPBY c.SubscriberKey, c.EmailAddress, c.FirstName, c.LastName
HAVINGCOUNT(DISTINCT e.EventDate) >=3
Configuration
Set up environment variables:
export GENPARK_API_KEY=your_api_key_here
export SFMC_SCHEMA=your_schema_name # Optional: default schema for data extensionsexport GENPARK_MODEL=gpt-4 # Optional: specify LLM model
from genpark_audience_generator import AudienceGenerator
generator = AudienceGenerator(api_key=os.getenv('GENPARK_API_KEY'))
# High-value customers based on purchase behavior
intent = """
Find customers who:
- Made purchases totaling over $1000 in the last 6 months
- Have purchased at least 3 times
- Are subscribed to the premium newsletter
"""
sql = generator.generate_sql(intent)
print(sql)
Re-engagement Campaign
# Identify customers to re-engage
intent = """
Target customers who:
- Haven't purchased in 60-90 days
- Previously purchased more than twice
- Have opened emails in the last 30 days
- Are not unsubscribed
"""
sql = generator.generate_sql(
intent,
output_fields=['SubscriberKey', 'EmailAddress', 'LastPurchaseDate']
)
Geographic Targeting
# Location-based segment
intent = """
Customers in California or New York
who have purchased outdoor gear
and have a lifetime value over $500
"""
sql = generator.generate_sql(intent, schema='ENT.RetailCustomers')
Advanced Usage
Custom Data Extension Mapping
from genpark_audience_generator import AudienceGenerator, DataExtensionConfig
# Define your data extension schema
de_config = DataExtensionConfig(
name='CustomersDE',
fields={
'SubscriberKey': 'Text',
'EmailAddress': 'EmailAddress',
'FirstName': 'Text',
'LastName': 'Text',
'City': 'Text',
'State': 'Text',
'LTV': 'Decimal',
'LastPurchase': 'Date'
},
primary_key='SubscriberKey'
)
generator = AudienceGenerator(
api_key=os.getenv('GENPARK_API_KEY'),
data_extensions=[de_config]
)
intent = "Customers in Texas with LTV over $1000"
sql = generator.generate_sql(intent, target_de='CustomersDE')
Batch Processing
import json
from genpark_audience_generator import AudienceGenerator
generator = AudienceGenerator(api_key=os.getenv('GENPARK_API_KEY'))
# Process multiple audience definitions
audience_intents = [
"Active customers in the last 30 days",
"Abandoned cart in the last 7 days",
"VIP customers with 10+ purchases",
"Newsletter subscribers who never purchased"
]
results = []
for intent in audience_intents:
try:
sql = generator.generate_sql(intent)
results.append({
'intent': intent,
'sql': sql,
'status': 'success'
})
except Exception as e:
results.append({
'intent': intent,
'error': str(e),
'status': 'failed'
})
# Save resultswithopen('audience_queries.json', 'w') as f:
json.dump(results, f, indent=2)
Query Validation
from genpark_audience_generator import AudienceGenerator, QueryValidator
generator = AudienceGenerator(api_key=os.getenv('GENPARK_API_KEY'))
validator = QueryValidator()
intent = "Customers who clicked on Black Friday email"
sql = generator.generate_sql(intent)
# Validate before using in Marketing Cloud
validation = validator.validate(sql)
if validation.is_valid:
print(f"Valid SQL: {sql}")
else:
print(f"Validation errors: {validation.errors}")
API Reference
AudienceGenerator Class
classAudienceGenerator:
def__init__(
self,
api_key: str,
endpoint: str = None,
data_extensions: list = None,
model: str = 'gpt-4'):
"""Initialize the audience generator"""passdefgenerate_sql(
self,
intent: str,
schema: str = None,
target_de: str = None,
output_fields: list = None,
validate: bool = True) -> str:
"""Generate Marketing Cloud SQL from natural language"""passdefexplain_query(self, sql: str) -> dict:
"""Get natural language explanation of SQL query"""pass
Key Methods
generate_sql(): Converts natural language to SQL
explain_query(): Reverse operation - SQL to natural language
validate_schema(): Check data extension schema compatibility
optimize_query(): Suggest query optimizations
CLI Usage
# Generate SQL from command line
python genpark_audience_generator.py --intent "customers who purchased last week"# With custom schema
python genpark_audience_generator.py \
--intent "high-value customers" \
--schema "ENT.RetailCustomers" \
--output output.sql
# Batch mode
python genpark_audience_generator.py --batch intents.txt --output-dir ./queries/
# Validate existing query
python genpark_audience_generator.py --validate existing_query.sql
Example: Complete Workflow
import os
from genpark_audience_generator import AudienceGenerator
from sfmc_client import MarketingCloudClient # Hypothetical SFMC client# Initialize
generator = AudienceGenerator(api_key=os.getenv('GENPARK_API_KEY'))
mc_client = MarketingCloudClient(
client_id=os.getenv('SFMC_CLIENT_ID'),
client_secret=os.getenv('SFMC_CLIENT_SECRET')
)
# Define audience intent
intent = """
Create a segment of customers who:
- Purchased women's apparel in Q4 2025
- Have an email open rate above 25%
- Live in metro areas
- Are not currently in any active campaign
"""# Generate SQL
sql_query = generator.generate_sql(
intent,
schema='ENT.CustomerMaster',
output_fields=['SubscriberKey', 'EmailAddress', 'FirstName', 'City']
)
# Create data extension and execute query
de_name = 'Q4_WomensApparel_Engaged_Metro'
mc_client.create_data_extension(de_name, sql_query)
mc_client.execute_query(sql_query, target_de=de_name)
print(f"Audience segment created: {de_name}")
Troubleshooting
API Authentication Errors
# Check API key configurationimport os
ifnot os.getenv('GENPARK_API_KEY'):
raise ValueError("GENPARK_API_KEY environment variable not set")
# Test connectionfrom genpark_audience_generator import AudienceGenerator
generator = AudienceGenerator(api_key=os.getenv('GENPARK_API_KEY'))
generator.test_connection()
Invalid SQL Generated
# Enable debug mode for detailed logging
generator = AudienceGenerator(
api_key=os.getenv('GENPARK_API_KEY'),
debug=True
)
# Add more context to intent
intent = """
Data extension: Customers (fields: SubscriberKey, EmailAddress, City, State, LTV)
Find: Customers in California with LTV > 1000
"""
sql = generator.generate_sql(intent)
