Struct QualitySignals 

pub struct QualitySignals {
    pub tool_error_rate: f32,
    pub iterations_used: usize,
    pub max_iterations: usize,
    pub was_truncated: bool,
    pub has_failure_patterns: bool,
    pub is_empty_response: bool,
}

Heuristic quality signals extracted from an agent response.

These signals drive the quality gate (ERL’s τ threshold) that determines whether a reflection should be triggered.

Fields

tool_error_rate: f32

Fraction of tool calls that resulted in errors (0.0–1.0).

iterations_used: usize

Number of agentic loop iterations used.

max_iterations: usize

Maximum iterations configured.

was_truncated: bool

Whether the response was truncated due to token limits.

has_failure_patterns: bool

Whether the response contains apology/failure patterns.

is_empty_response: bool

Whether the response is empty or near-empty.

Implementations

impl QualitySignals

pub fn score(&self) -> f32

Compute an aggregate quality score (0.0–1.0, higher is better).

This is the heuristic that replaces ERL’s verifiable reward signal. In our setting we don’t have a ground-truth reward, so we use observable proxy signals from the agent’s execution.

Trait Implementations

impl Clone for QualitySignals

fn clone(&self) -> QualitySignals

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for QualitySignals

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

Formats the value using the given formatter.