Generation Agents

The five agents that form the test generation pipeline — from crawling your live app to producing reviewed, ready-to-run pytest scripts.

For a detailed description of how these agents interact as a pipeline, see Generation Pipeline. This page documents each agent's individual behaviour and configuration.

AppExplorer

Purpose: Crawls the live application and captures selector and structure data for every page or endpoint referenced in the Knowledge Base.

What it does:

• Launches a headless Chromium browser and navigates to each page listed in KB domain base URLs
• Discovers interactive elements (inputs, buttons, links, forms, selects)
• Records multiple selector strategies per element (data-testid, ARIA role, text, CSS, XPath)
• For API modules: probes API endpoints and records response shapes
• Stores output as structured JSON selector maps per domain

Re-run AppExplorer only: Navigate to Project → Generation → AppExplorer → Re-run after UI changes to refresh selector maps without regenerating scripts.

DocGen

Purpose: Converts KB scenarios into structured test case documents that serve as the specification for ScriptGen.

What it does:

• Reads each KB scenario (steps, expected outcomes, priority, tags)
• Combines with AppExplorer selector data to infer element interactions
• Generates a Markdown test case document with explicit test setup, actions, and assertions
• Assigns unique test case IDs following the configured numbering convention

DocReview

Purpose: Quality-gates DocGen output before code is written, catching gaps and ambiguities early.

Checks performed:

• All required document fields are present and non-empty
• Steps and expected outcomes are internally consistent
• Referenced DOM selectors exist in AppExplorer output
• Assertions are specific enough for ScriptGen to produce meaningful code

Passing documents are forwarded to ScriptGen. Failing documents are retried once automatically; persistent failures enter the Approvals queue as rejected items with the specific failure reason.

ScriptGen

Purpose: Converts approved test case documents into executable pytest test scripts.

What it produces:

• Python pytest files using Playwright (Web UI) or HTTPX (API) as the testing library
• One test function per scenario ID
• Fixtures for authentication, environment configuration, and base URL injection
• Assertions derived from the expected outcomes in the test case document

ScriptReview

Purpose: Validates generated scripts before they reach the Approvals queue.

Checks performed:

• Python syntax validity — scripts are parsed and any syntax errors are reported
• Selector existence — verifies referenced locators exist in current AppExplorer maps
• Assertion presence — each test function must contain at least one assertion
• Import validity — checks that all used fixtures and imports are resolvable

Passing scripts are sent to the Approvals queue. Failing scripts trigger one automatic ScriptGen retry; persistent failures go to Approvals as rejected items.