Design thoughtful engineering approach

[LopeWale]

…ck testing

This commit introduces the core AI intelligence layer for TestAble, enabling true LLM-powered test automation with both frontend and backend testing capabilities.

New Components (backend/ai/)

engine.py - Core AI Engine

• Proper Stagehand SDK integration using page.act(), page.observe(), page.extract()
• Cache-first strategy with intelligent fallback to AI
• Action execution with full metrics and reasoning
• Async context manager for clean resource management

action_parser.py - Natural Language Parser

• Comprehensive NL instruction parsing (20+ action types)
• Element type detection (button, link, input, etc.)
• Value extraction from quoted strings
• Position extraction (first, second, last)
• Container context extraction (within forms, modals)
• Confidence scoring for ambiguous instructions

assertions.py - AI-Powered Assertions

• Semantic assertion verification using AI understanding
• Multiple assertion types: visibility, text, value, count
• Element state verification with reasoning
• Fluent expectation builder API
• Fallback text verification for robustness

api_tester.py - Backend API Testing

• Full HTTP client with request/response handling
• JSON schema validation
• Content verification (partial matching)
• CRUD operation testing helper
• Test suite execution with aggregation

test_runner.py - Test Orchestration

• Unified runner for frontend and backend tests
• Natural language test step parsing
• TestCase fluent builder API
• Suite execution with parallel support
• Result aggregation and metrics

Updated Components

testable_client.py

• Proper Stagehand AI integration using page.act()
• Intelligent selector matching as fallback
• Integration with new ActionParser module

Tests

• Comprehensive test suite (16 passing tests)
• Tests for all parser action types
• Tests for type inference and configuration
• Integration tests for full flow

This architecture enables TestAble to accurately test both frontend UI components and backend APIs, with AI-powered understanding of natural language test instructions.