Problem
Key decision parameters are hardcoded in decision.py:
schema_break_rate hard stop at 0.25
high_criticality_failure_rate hard stop at 0.20
proceed_if_overall_risk_below at 0.08
- High-criticality weight at 2.0x
- Latency thresholds (p50: 1.5x, p95: 2.0x)
Users with different risk tolerances (e.g., internal tools vs. customer-facing APIs) can't tune these without editing source code.
Proposal
Add an optional decision section to the YAML config:
decision:
hard_stop_schema_break_rate: 0.25
hard_stop_high_crit_failure_rate: 0.20
proceed_risk_threshold: 0.08
high_criticality_weight: 2.0
latency_p50_threshold: 1.5
latency_p95_threshold: 2.0
All fields optional with current values as defaults — zero breaking changes.
Impact
- Users can tune sensitivity per migration
- Enables stricter thresholds for critical systems, looser for experimentation
Problem
Key decision parameters are hardcoded in
decision.py:schema_break_ratehard stop at 0.25high_criticality_failure_ratehard stop at 0.20proceed_if_overall_risk_belowat 0.08Users with different risk tolerances (e.g., internal tools vs. customer-facing APIs) can't tune these without editing source code.
Proposal
Add an optional
decisionsection to the YAML config:All fields optional with current values as defaults — zero breaking changes.
Impact