Module result_parser 

Standardized tool result parsing (R-AS-01).

Diverse security detectors (classifiers, heuristics, LLM judges, etc.) produce wildly different output formats. This module normalises those outputs into a unified DetectorResult schema, then aggregates multiple detector results into a single ScanResult using pluggable AggregationStrategy policies.

Quick start

use llmtrace_security::result_parser::*;
use llmtrace_core::SecuritySeverity;

let r1 = DetectorResult::new("injecguard", DetectorType::Classifier)
    .with_threat(ThreatCategory::InjectionDirect)
    .with_confidence(0.92)
    .with_severity(SecuritySeverity::High);

let r2 = DetectorResult::new("heuristic-v1", DetectorType::Heuristic)
    .with_threat(ThreatCategory::Benign)
    .with_confidence(0.85)
    .with_severity(SecuritySeverity::Info);

let aggregator = ResultAggregator::new(AggregationStrategy::MajorityVote);
let agg = aggregator.aggregate(&[r1, r2]);

Structs

AggregatedResult

The outcome of combining multiple detector results.

DetectorResult

Normalised output from a single security detector.

ResultAggregator

Combines multiple DetectorResults into a single verdict.

ScanResult

Final scan output combining all detector results and their aggregation.

ScanResultBuilder

Incrementally builds a ScanResult from individual detector results.

Enums

AggregationStrategy

Strategy for combining multiple detector results.

DetectorType

Kind of security detector that produced a result.

ThreatCategory

Threat classification for a scanned input.

Functions

compute_input_hash

Deterministic hash of input text. Uses a simple multiplicative hash folded to 16 hex characters.