Smart Test Dependency Caching & Discovery V3

🚀 Smart Test Dependency Caching

Tests with dependencies (like login flows) just got dramatically faster. Our new caching system stores the results of dependency execution—including storage state, variables, and final URLs—so dependent tests can skip re-running unchanged prerequisites.

What's New:

• Transitive Dependency Intelligence: The system now tracks full dependency chains (A→B→C→D) and loads the deepest cached state, executing only what's changed downstream
• Hash-Based Invalidation: Caches automatically invalidate when test definitions, variables, or dependencies change—no stale results, ever
• Storage State Restoration: Cached dependencies restore browser state (cookies, localStorage) exactly where they left off, eliminating redundant navigation
• Final URL Navigation: Tests resume from the exact URL where cached dependencies completed, maintaining proper execution flow
• Variable Propagation: Cached variables flow correctly through dependency chains, ensuring consistent test data

Business Impact: For test suites with shared login flows or multi-step prerequisites, this means 60-80% faster execution times on subsequent runs. Your CI/CD pipelines complete faster, developers get feedback sooner, and you ship with confidence at velocity.

🧠 Discovery V3: Two-Phase Intelligent Test Discovery

We've completely redesigned our autonomous test discovery workflow with a two-phase architecture that separates analysis from test creation. The result? Smarter test coverage with 80-90% fewer RAG queries.

How It Works:

• Phase 1 - Analysis: A dedicated analysis agent systematically explores your app (2-3 levels deep), performs ONE comprehensive search of existing tests, and builds a user stories registry with embeddings for semantic search
• Phase 2 - Test Creation: Armed with full context from Phase 1, the creation agent identifies coverage gaps and spawns targeted tests with proper dependency orchestration (automatic login test detection)
• Persistent Context: User stories are stored in KB storage with embeddings, enabling semantic queries like "find stories about checkout flow" without repeated crawling

ROI for QA Leaders: Discovery V3 eliminates redundant exploration and dramatically reduces LLM costs while producing better-targeted test coverage. Your QA team spends less time on manual test creation and more time on strategic quality initiatives.

🛠️ Infrastructure & Developer Experience

We've also shipped foundational improvements that make desplega.ai more powerful and easier to use:

• CLI Test Runner: New cli-test command for running E2E tests locally from YAML files, with screenshot capture, progress streaming, and configuration management
• Refactored Test Execution: Core execution logic now shared between Hatchet workflows and CLI, with cleaner separation of concerns and comprehensive test coverage
• Deterministic Conversion Persistence: AI-to-Playwright code conversions now save correctly even on re-runs, reducing AI invocations and improving test stability
• Flaky Test Reporting Fix: Daily summary notifications now correctly distinguish between real failures and flaky tests (failed initially but passed on retry), giving you accurate quality metrics
• React Best Practices: Frontend optimizations including immutable array operations (toSorted()), hoisted RegExp patterns, and O(1) Set lookups for better performance

Performance & Quality Impact

These improvements align directly with DORA metrics and shift-left testing principles:

• Deployment Frequency: Faster test execution means faster CI/CD cycles
• Lead Time for Changes: Reduced test suite runtime accelerates feedback loops
• Change Failure Rate: Better test coverage from Discovery V3 catches bugs earlier
• Cost Efficiency: Smart caching and reduced RAG queries lower testing costs