Struct Validator 

pub struct Validator {
    pub limits: ValidationLimits,
}

Validates AgentOutput fields against configurable limits.

Uses clean_title for normalizing finding titles before length validation, and configurable limits for field lengths and counts.

Fields

limits: ValidationLimits

Active validation limits.

Implementations

impl Validator

pub fn new() -> Self

Creates a validator with default limits.

pub fn with_limits(limits: ValidationLimits) -> Self

Creates a validator with custom limits.

pub fn validate(&self, output: &AgentOutput) -> Result<(), MagiError>

Validates an AgentOutput, returning on first failure.

Checks in order: confidence, summary, reasoning, recommendation, findings. Returns MagiError::Validation with a descriptive message on failure.

pub fn validate_mut(&self, output: &mut AgentOutput) -> Result<(), MagiError>

Validates output in place, replacing each finding’s title with its cleaned form (see clean_title) before length validation.

This is the preferred entry point for pipelines that parse LLM responses, because it ensures downstream code sees titles in the canonical cleaned form used by the consensus engine.

Atomicity

This method is atomic on error: it cleans and validates all finding titles into a temporary buffer before committing any mutation to output. If any validation check fails, output is left completely unchanged — no partial title cleaning occurs.

Errors

Returns MagiError::Validation on the first field that fails validation. Validation order: confidence → summary → reasoning → recommendation → findings (count, then each cleaned title/detail). On error, output is not modified.

Trait Implementations

impl Default for Validator

fn default() -> Self

Returns the “default value” for a type.