Examples¶
Complete working examples for different ACGP implementations.
By Conformance Level¶
-
Minimal
Minimal ACGP implementation for development and testing
-
Standard
Production-ready implementation with full features
-
Complete
Mission-critical implementation with advanced security
By Use Case¶
Customer Service Agent¶
Simple customer service bot with refund capabilities
from acgp import GovernanceSteward, CognitiveTrace
steward = GovernanceSteward(
acl_tier="ACL-2",
blueprint="customer-service",
tripwires={"max_refund": 500}
)
# Agent decides to issue refund
trace = CognitiveTrace(
reasoning="Customer complaint is valid, product defective",
action="issue_refund",
parameters={"amount": 100}
)
result = steward.evaluate(trace)
if result.intervention == "OK":
process_refund(100)
Financial Trading Agent¶
High-stakes trading agent with strict oversight
steward = GovernanceSteward(
acl_tier="ACL-4",
blueprint="financial-trading",
tripwires={
"max_transaction": 10000,
"daily_limit": 50000
}
)
Data Processing Agent¶
Batch data processor with moderate autonomy
steward = GovernanceSteward(
acl_tier="ACL-2",
blueprint="data-processing",
tripwires={
"max_records": 10000,
"max_modifications": 1000
}
)
By Framework¶
- LangChain - See Integration Guide
- Autogen - See Integration Guide
- Custom Agents - See Integration Guide