@vijaypjavvadi
Open-Source npm Packages
Eight published packages from the research program. Version and download badges are fetched live from npm; cumulative totals below are as of October 2026.
@vijaypjavvadi/bdd2pw
Gherkin → Playwright scaffolder. Generates runnable Playwright TypeScript tests from .feature files, detects existing Page Objects, inspects live pages via the Microsoft Playwright MCP, and emits POMs and specs ready for execution. CLI + HTTP service with optional governed LLM fallback. Described in a peer-reviewed article in Elsevier SoftwareX.
@vijaypjavvadi/sel2pw
Selenium → Playwright migration toolkit. Deterministic AST- and rule-based conversion of Java Selenium suites (TestNG, JUnit, Cucumber) to Playwright TypeScript scaffolds, with experimental C# NUnit and SpecFlow support and an optional LLM fallback. Publishes via GitHub Actions OIDC.
@vijaypjavvadi/pw-emit
Shared emitter library that renders Playwright TypeScript Page Objects, spec files, and project scaffolds from a source-agnostic intermediate representation. Powers both bdd2pw and sel2pw.
@vijaypjavvadi/usagecontract
Usage-aware consumer-driven contract testing for HTTP APIs. Derives per-consumer contracts from observed traffic and weights verification by real usage, so breaking changes are caught where they matter most.
@vijaypjavvadi/pw-self-heal
In-process self-healing locators for Playwright. A 17-feature ONNX model ranks candidate elements when a locator breaks, with heuristic, ML, and hybrid modes and confidence-gated heal-vs-flag decisions. Companion to the self-healing study under review at Wiley STVR.
@vijaypjavvadi/synthdata
LLM-driven synthetic test-data generation with referential integrity across related tables and schemas. Companion to the test-data memorisation study under review at Springer Empirical Software Engineering.
@vijaypjavvadi/agentic-qa
Standalone agentic QA runtime: an LLM selects and invokes tools over MCP inside a governance gate, with a replayable ledger so every decision is auditable and reproducible. Substrate for the test-agent weakening experiment in design.
Open Research Datasets
Datasets and reproducibility bundles archived on Zenodo with persistent DOIs. Each one underpins at least one manuscript on the Publications page.
Software Defect Prediction Dataset
296,457 file-level instances mined from five mature open-source systems (Elasticsearch, Spring Boot, Hadoop, Kafka, Express) with process metrics and defect labels. Anchors the defect-prediction, cross-repository transfer, and risk-based testing manuscripts.
Self-Healing Automation Dataset
11,100 rows covering 2,400 UI mutation events across seven refactoring classes, with the eight locator feature families used to train the pw-self-heal ONNX model and evaluate tree-ensemble locator ranking.
RAITG Test Generator Dataset
312 natural-language requirements across three application domains (banking, healthcare, e-commerce) with verified executable tests — the benchmark corpus for LLM test-case generation studies.
TruthHire: Reference Implementation & Evaluation Harness
Deterministic timeline engine, adversarial AI-fake robustness harness, and public-corpus assessment (CC-BY-4.0) for the employment-fraud detection and privacy-preserving fraud-reporting manuscripts.
Behind the papers and the platform
Each project here is either the experimental pipeline behind a paper or a service running in the TestForge AI platform.
ai-defect-prediction-repository-analytics
The full experimental pipeline behind Papers A, B, and C — dataset construction, six-classifier benchmarking, LORO cross-repository transfer, and top-k risk-coverage analysis.
defect-attribution-service
FastAPI microservice hosting gb-paper1-v4-fixed_age. Exposes /predict and /coverage endpoints for risk-based test prioritisation.
test-case-generation-service
The RAITG production service: requirements in, executable Playwright + BDD tests out. Includes the deterministic rule-verifier and symbolic mutation-indicator adequacy check.
self_healing_stage_services
Self-healing runtime: DOM similarity, tree-ensemble locator ranking, mutation-event replay harness, and confidence-gated heal vs. flag decisions.
modern-automation-platform
Reference packaging and emit-layer for Playwright artefacts; the substrate the platform's higher-level services share.
ai-governance
The cross-cutting governance rail: payload controls, content redaction, author privacy hashing, prompt versioning, and verifiable audit trails on every model call.
test-prioritization-service
The CI/CD-facing service that maps risk scores to a deterministic execution order; supports advisory and gating rollouts.
testforge_ml
Shared ML utilities (calibration, SMOTE, threshold selection, feature engineering) used across all model-bearing services.
TruthHire-research
Reproducibility bundle for the employment-experience fraud-detection study: a deterministic timeline engine, an adversarial AI-fake robustness harness, and a public-corpus assessment (CC-BY-4.0). Archived at Zenodo 10.5281/zenodo.20681820.
TestForge PaperQC
Pre-submission QC for software-engineering manuscripts: venue fit, peer benchmarking, structure, honesty & integrity checks, and a submission-package recompile. Free during beta.