KO
EN
JAIn Agile and Cloud-Native environments where software deployment cycles are shortened from 'weekly' to 'daily', Quality Assurance (QA) remains the area with the most manual work and resource bottlenecks. The 'Maintenance Hell', where existing E2E test scripts easily break even with minor UI changes (DOM structure, ID/Class attributes, etc.), has been the biggest obstacle to automation.
To overcome these limitations, Empasy's SyncETA has built an End-to-End Self-Evolving AI First QA Platform encompassing everything from AI automated generation to unmanned autonomous operations, under the vision of "Boon to business by agility". This article delves into the core architecture of SyncETA and its latest engineering advancements.
1. Intelligent 5-Layer Architecture Data Flow
SyncETA consists of five core hierarchical structures that evolve themselves by precisely collecting unstructured user behavior logs.
Core Differentiator: Architecture Saving 99% of Annual LLM Costs
Unlike existing AI tools, SyncETA utilizes the LLM 'only once at the time of scenario generation', and then executes through a deterministic record playback engine without LLM API calls during the actual test execution phase. This innovatively reduces the cost of regression tests repeated thousands of times in an enterprise environment by up to 99%.
2. Key Engineering Advancements
① Headless Automation Pipeline & External System Integration (RESTful API)
Provides a perfect unmanned automation loop that immediately completes tests in the background just by sending natural language command text from an external system or CI/CD pipeline.
- Inbound Interface (POST
/api/v1/execute/natural-language): When receiving a project ID and a natural language scenario from the outside, the embedded LLM engine analyzes the context and converts it in real time into Playwright-MCP commands structured asLogin→Navigate→Verify Element. - Automatic Video Evidence Mapping: Runs at high speed in parallel in Headless mode without displaying the browser, records the entire execution process in real time in
.webmformat, automatically uploads it to an AWS S3 Bucket after the test ends, and immediately returns the unique Public URL and Pass/Fail status log to the external Callback URL (Outbound API).
3. Enterprise Adoption Performance & Future Value
- 80% Reduction in QA Design & Writing Effort: Because communication is possible with natural language Excel forms in an international standard format, it provides a 'QA Democratization' environment where even business personnel and planners (PL) without coding knowledge can immediately design and verify quality assets.
- Security-First Local LLM Ecosystem: To prevent the risk of leaking sensitive corporate data or test environment information, it fully supports the construction of Local LLM models with a Docker-based completely closed-network On-premise and isolated cloud hybrid infrastructure architecture.
Based on the precise verification performance of the web ecosystem, SyncETA is realizing the technology development milestone of 'Mobile Domain Expansion (DeviceHub Integration)' based on the Universal Locator structure that can control Android and iOS simultaneously with a single natural language scenario. SyncETA's advanced new technology that will change the QA paradigm will be continuously shared through the Empasy official tech blog.