Skip to content

Cross-Fact Rules

Prerequisites: Working Memory (fact storage).

Correlate multiple facts with cross-fact logic.

Single-fact limitation

Standard engines evaluate one fact at a time:

from fluxrules.engine.phreak import PhreakEngine

engine = PhreakEngine()
result1 = engine.evaluate({"customer_id": "c1", "amount": 5000})
result2 = engine.evaluate({"customer_id": "c2", "amount": 1000})
# No correlation between result1 and result2

Cross-fact correlation patterns

Pattern 1: Pre-aggregate upstream

Combine facts before evaluation:

# Combine multiple facts before rule evaluation
combined = {
    "transaction_1_amount": 5000,
    "transaction_2_amount": 3000,
    "total_24h": 8000,  # Pre-computed
    "customer_id": "c1"
}

result = engine.evaluate(combined)

Pattern 2: Multi-step with sessions

Use sessions to accumulate state:

# python skip
from fluxrules.services.rule_service import RuleService

service = RuleService.create()
session = service.create_session()

# Step 1: Check first condition
session.add_fact("transaction_1", {"amount": 5000})
result1 = session.evaluate()

# Step 2: Add second condition
session.add_fact("transaction_2", {"amount": 3000})
result2 = session.evaluate()

# Both facts now in session state

Pattern 3: Working memory

Assert multiple facts and query them:

# Store facts in working memory
id1 = engine.assert_fact({"customer_id": "c1", "amount": 5000})
id2 = engine.assert_fact({"customer_id": "c1", "amount": 3000})

# Evaluate combined logic
result = engine.evaluate({
    "customer_id": "c1",
    "total": 8000,  # Computed from stored facts
})

When to use

  • Multiple related facts → Aggregate and pass combined dict
  • Multi-step workflow → Use sessions with fact accumulation
  • Temporal correlation → Store facts in working memory

Next Steps