Technical Specifications →

Levox Technical Specifications

Concrete performance metrics, accuracy targets, detection stages, supported languages, and architecture overview.

285 files/sec
Average Scan Speed
benchmarks/reports/output/real_benchmarks.json
198 MB
Peak Memory Usage
benchmarks/reports/output/memory_results.json
81.8%
Precision (current)
benchmarks/reports/output/accuracy_results.json
4.3%
False Positive Rate
Ground-truth corpus

Detection Pipeline

  1. Regex Detection
  2. AST Analysis
  3. Context Analysis
  4. Dataflow Analysis
  5. CFG Analysis
  6. ML Filtering (XGBoost)
  7. GDPR/CCPA Compliance Mapping

Parsers & Languages

  • Python — full AST support
  • JavaScript/TypeScript — full AST support
  • Java — full AST support
  • Extensible Tree-sitter based architecture

Accuracy & ML

  • XGBoost-based false positive reduction
  • Pattern recognition and hotspot detection
  • Risk prediction and remediation effectiveness
  • Anomaly detection and explanations

Security

  • Offline scanning; no code leaves your environment
  • Cryptographic audit logs with SHA-256 hash chains
  • Role-based access & multi-tenant isolation
  • Configurable suppression and custom rules

Architecture Overview

  • Modular detection engines coordinated by a production Detection Engine
  • Evidence Package Generator for audit-ready reporting
  • Compliance engines for GDPR and CCPA with scoring and mapping
  • Realtime analytics via WebSocket server and Supabase integration
  • CI/CD integration suite with template generator and validator
  • Feature-based rate limiter with persistent storage
System requirements and detailed diagrams available upon request.