How aj-design19/testing_agent_updated works
Adya AI Testing Agent autonomously validates workflows by driving a browser-based SAI copilot, answering its questions as a selected persona, collecting evidence, and judging the run. Its wider capabilities include browser/canvas inspection, accessibility and visual evaluation, persistent learning, human review, and run reporting. The main orchestration excerpt is incomplete, so edges from it are limited to components named in the SOURCE INDEX; unshown subsystem wiring is omitted.
Main entry points: orchestrator.py Run orchestrator · canvas_monitor.py Canvas monitor · llm_judge.py LLM evaluation · models.py Persistence models
Components 26
Run orchestration
- Run entry
app/main.py - Run orchestrator
app/orchestrator.py
Browser and evidence
- Browser manager
app/browser/browser_manager.py - Login
app/browser/login.py - SAI interaction
app/browser/sai_handler.py - SAI navigation
app/browser/sai_navigator.py - Canvas monitor
app/browser/canvas_monitor.py - Flow validation
app/browser/flow_graph_validator.py - Canvas zoom
app/browser/canvas_zoomer.py - DOM intelligence
app/browser/dom_intelligence.py - Screenshot capture
app/browser/screenshot_agent.py - Accessibility audit
app/browser/accessibility_checker.py - Style extraction
app/browser/style_extractor.py
Workflow and conversation
- Workflow definitions
app/workflows/workflow_registry.py - Journey loader
app/workflows/journey_loader.py - Persona answers
app/conversation/answer_engine.py
Evaluation and reporting
- LLM evaluation
app/judge/llm_judge.py - Visual evaluation
app/judge/visual_judge.py - Run report
app/reporting/run_report.py
Learning and governance
- Persistence models
app/store/models.py - Mongo store
app/store/mongo_store.py - Postgres store
app/store/postgres_store.py - Audit log
app/store/audit_log.py - Self-learning service
app/esll/esll_service.py - Human review queue
app/governance/hitl_manager.py
Other components
- Test operator
Connections 35
- Test operator → Run entry: starts run
- Run entry → Run orchestrator: dispatches to
app/main.py - Run orchestrator → Browser manager: uses browser
app/orchestrator.py - Run orchestrator → Login: uses login
app/orchestrator.py - Run orchestrator → SAI interaction: drives interaction
app/orchestrator.py - Run orchestrator → Screenshot capture: captures evidence
app/orchestrator.py - Run orchestrator → Canvas monitor: monitors canvas
app/orchestrator.py - Run orchestrator → Accessibility audit: checks accessibility
app/orchestrator.py - Run orchestrator → Style extraction: checks styling
app/orchestrator.py - Run orchestrator → Persona answers: uses answers
app/orchestrator.py - Run orchestrator → LLM evaluation: evaluates run
app/orchestrator.py - Journey loader → Workflow definitions: loads definitions
app/workflows/journey_loader.py - SAI interaction → Persona answers: generates replies
app/browser/sai_handler.py - SAI interaction → SAI navigation: reads panel
app/browser/sai_handler.py - SAI interaction → DOM intelligence: reads DOM state
app/browser/sai_handler.py - Canvas monitor → DOM intelligence: inspects DOM
app/browser/canvas_monitor.py - Canvas monitor → SAI navigation: navigates SAI
app/browser/canvas_monitor.py - Canvas monitor → Screenshot capture: uses screenshots
app/browser/canvas_monitor.py - Canvas monitor → Flow validation: validates flow
app/browser/canvas_monitor.py - Canvas monitor → Canvas zoom: zooms canvas
app/browser/canvas_monitor.py - LLM evaluation → Visual evaluation: uses visual judge
app/judge/llm_judge.py - LLM evaluation → Flow validation: uses flow checks
app/judge/llm_judge.py - LLM evaluation → Accessibility audit: uses audit results
app/judge/llm_judge.py - LLM evaluation → Style extraction: uses style results
app/judge/llm_judge.py - Run report → Visual evaluation: renders evaluations
app/reporting/run_report.py - Mongo store → Persistence models: uses models
app/store/mongo_store.py - Postgres store → Persistence models: uses models
app/store/postgres_store.py - Audit log → Persistence models: uses models
app/store/audit_log.py - Self-learning service → Persistence models: uses models
app/esll/esll_service.py - Self-learning service → Mongo store: persists learning
app/esll/esll_service.py - Self-learning service → Postgres store: persists runs
app/esll/esll_service.py - Self-learning service → Audit log: records changes
app/esll/esll_service.py - Human review queue → Postgres store: manages queue
app/governance/hitl_manager.py - Human review queue → Audit log: audits decisions
app/governance/hitl_manager.py - Human review queue → Persistence models: uses queue models
app/governance/hitl_manager.py