Struct SafetyAnalyzer 

pub struct SafetyAnalyzer { /* private fields */ }

Safety analyzer for detecting harmful, toxic, or inappropriate content

Implementations

impl SafetyAnalyzer

Rule-based, keyword-heuristic implementations. See the module docs and HealthTracker: every score below is a deterministic function of the literal response text (there is no trained classifier or knowledge base behind this), and confidence is always derived from a real, measurable property of the match (never a flat literal).

pub fn new(_config: &LLMDebugConfig) -> Self

pub async fn analyze_safety( &mut self, response: &str, ) -> Result<SafetyAnalysisResult>

pub fn get_health_summary(&self) -> HealthSummary

Real health summary derived from Self::health’s running history of Self::analyze_safety calls – status and trend used to be hardcoded (trend via a safety_trend field that was set once at construction and never updated).

Trait Implementations

impl Debug for SafetyAnalyzer

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.