@happyvertical/smrt-support 0.51.30 → 0.51.32

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -76,6 +76,16 @@ The AI workflow fails toward a human. Client requests, low confidence, high
76
76
  severity, sensitive categories, failed resolution, and policy limits always
77
77
  remain handoff triggers.
78
78
 
79
+ When typed decisions are configured, the default classification boundary uses
80
+ the case plan's ordered severity definitions and evaluates sensitivity as a
81
+ predicate. `SupportPolicy.allowedCategories` is optional: supplying values
82
+ turns category into a closed decision vocabulary; leaving it empty preserves
83
+ free-category generation. A configured ambiguous severity/category stays
84
+ unassigned and hands off to a human; an exactly ambiguous sensitivity predicate
85
+ does so too. The recorded classification confidence is an operating aggregate for
86
+ audit only, never a calibrated answer confidence or the policy answer threshold.
87
+ App-injected `SupportAiBoundary` implementations remain authoritative.
88
+
79
89
  ## Commercial separation
80
90
 
81
91
  Managed Support Plans define client coverage, targets, included time, and