Autonomous Quality Agent
AI-Powered API Testing Platform + Rust Engine
About the Project
AQA is a quality engineering platform that acts as an intelligent agent. It automatically transforms requirements into executable tests using AI (LLM), combining Python cognition with high-performance Rust execution. The system follows a two-component decoupled architecture: the Brain (Python) for planning and the Runner (Rust) for massive parallel execution.
Brain + Runner Architecture
The project follows an innovative two-component architecture communicating via UTDL (Universal Test Definition Language) protocol:
🧠The Brain (Python 3.11+): Responsible for cognition, planning and validation. Reads requirements from technical documentation (Swagger/OpenAPI), generates test plans via LLM and validates against UTDL schema.
🦀 The Runner (Rust + Tokio): Deterministic and high-performance execution engine. Consumes UTDL plans, executes HTTP requests with massive parallelism via DAG (Directed Acyclic Graph).
AI-Powered Test Generation
The system ingests OpenAPI/Swagger specifications and automatically detects authentication schemes (API Key, Bearer, OAuth2).
Automatically generates negative test cases (400, 401, 404, 500) and creates complete authentication flows. All through optimized prompts for GPT-4.
Rigorous validation includes circular dependency detection, JSONPath validation and assertion type verification.
DAG-Based Parallel Execution
The Rust Runner implements a Directed Acyclic Graph (DAG) based execution engine. Steps with dependencies execute in order, while independent steps execute in parallel.
Supports 20+ simultaneous branches with configurable retry policies. JSONPath data extraction enables variable interpolation between steps.
Native integration with OpenTelemetry for distributed tracing and complete observability.
Professional CLI (10 Commands)
The CLI built with Click + Rich offers 10 modular commands following the Registry Pattern:
init - Initialize workspace | generate - Generate plans via LLM | validate - Validate UTDL plans | run - Execute tests
plan-list - List plans | config - Manage configuration | storage - Storage backend | cache - Cache management | trace - Tracing and telemetry
Stack & Design Patterns
The project uses enterprise design patterns: Strategy Pattern for LLM Providers, Factory Pattern for Storage Backends, Registry Pattern for CLI commands, Adapter Pattern for UTDL normalization.
518 tests cover Unit, Integration, E2E, Extreme and Security Audit scenarios. 13 specific security tests ensure credentials never leak to the LLM.