Documentation
¶
Overview ¶
Package reflex models condition-triggered safety responses.
Reflexes are trained responses: condition→action pairs that fire when specific conditions are met. Policy rules are reflexes. Rate limiting is a reflex. Escalation logic is a reflex. They are the Shield equivalent of Cortex's Behaviors.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Action ¶
type Action struct {
Type ActionType `json:"type"`
Target string `json:"target,omitempty"` // what to act on
Value any `json:"value,omitempty"` // action-specific value
Fallback string `json:"fallback,omitempty"` // fallback response text
}
Action defines what a reflex does when triggered.
type ActionType ¶
type ActionType string
ActionType defines what happens when a reflex triggers.
const ( // ActionBlock prevents content from passing. ActionBlock ActionType = "block" // ActionRedact replaces sensitive content with placeholders. ActionRedact ActionType = "redact" // ActionFlag marks content for review. ActionFlag ActionType = "flag" // ActionRewrite transforms content. ActionRewrite ActionType = "rewrite" // ActionEscalate sends content for human review. ActionEscalate ActionType = "escalate" // ActionLog records the event without modifying content. ActionLog ActionType = "log" // ActionThrottle rate-limits the request. ActionThrottle ActionType = "throttle" )
type ListFilter ¶
ListFilter defines filtering options for listing reflexes.
type Reflex ¶
type Reflex struct {
shield.Entity
ID id.ReflexID `json:"id" bun:",pk"`
Name string `json:"name" bun:",notnull"`
Description string `json:"description,omitempty"`
AppID string `json:"app_id" bun:",notnull"`
TenantID string `json:"tenant_id,omitempty"`
Triggers []Trigger `json:"triggers,omitempty" bun:"type:jsonb"`
Actions []Action `json:"actions,omitempty" bun:"type:jsonb"`
Priority int `json:"priority,omitempty" bun:",default:0"`
Enabled bool `json:"enabled" bun:",notnull,default:true"`
Metadata map[string]any `json:"metadata,omitempty" bun:"type:jsonb"`
}
Reflex represents a condition-triggered safety response. Reflexes are the sixth layer of the safety engine. They evaluate policy rules and custom conditions to produce actions.
type Store ¶
type Store interface {
CreateReflex(ctx context.Context, r *Reflex) error
GetReflex(ctx context.Context, rID id.ReflexID) (*Reflex, error)
GetReflexByName(ctx context.Context, appID, name string) (*Reflex, error)
UpdateReflex(ctx context.Context, r *Reflex) error
DeleteReflex(ctx context.Context, rID id.ReflexID) error
ListReflexes(ctx context.Context, filter *ListFilter) ([]*Reflex, error)
}
Store defines persistence operations for reflexes.
type Trigger ¶
type Trigger struct {
Type TriggerType `json:"type"`
Pattern string `json:"pattern,omitempty"` // regex/pattern for matching
Threshold float64 `json:"threshold,omitempty"` // score threshold
Window string `json:"window,omitempty"` // time window for rate triggers (e.g., "1m", "5m")
}
Trigger defines when a reflex activates.
type TriggerType ¶
type TriggerType string
TriggerType defines when a reflex activates.
const ( // TriggerOnScore activates when a check score exceeds a threshold. TriggerOnScore TriggerType = "on_score" // TriggerOnFinding activates when a specific finding type appears. TriggerOnFinding TriggerType = "on_finding" // TriggerOnPattern activates when text matches a pattern. TriggerOnPattern TriggerType = "on_pattern" // TriggerOnContext activates when context metadata matches. TriggerOnContext TriggerType = "on_context" // TriggerOnRate activates on rate-limit conditions. TriggerOnRate TriggerType = "on_rate" // TriggerAlways activates unconditionally. TriggerAlways TriggerType = "always" )
Source Files