🤝 Top Collaborators
.png&w=640&q=75)
.png&w=640&q=75)
🤓 Latest Submissions
TestForge
Test coverage on critical code paths is often low, but teams rarely know which untested functions actually matter most — writing tests manually is slow, so the riskiest, most-changed code frequently stays unverified until it breaks in production. TestForge solves this by combining static analysis with IBM Bob's agentic capabilities. It scans a target codebase, computes a risk score per function (cyclomatic complexity weighted by git commit churn), and ranks untested code by how urgently it needs coverage. For the highest-risk functions, it dispatches parallel calls to Bob in agent mode — each call analyzes a single function's source and logic, then writes a complete, real Jest test file addressing normal cases, edge cases, and error paths. Generated tests are never hand-templated boilerplate; they reflect actual reasoning about each function's behavior. Every generated test is then executed automatically. Passing tests are kept, failures are flagged rather than silently claimed as coverage. Coverage is re-measured and diffed against the original baseline, producing a concrete, honest before/after number — not an estimate. A web dashboard shows the full pipeline live: ranked risk table, per-file generation status (which used Bob vs. a fallback), pass/fail counts, and the resulting coverage delta. Target users are engineering teams who know coverage matters but don't have the bandwidth to prioritize and write tests manually — TestForge turns "we should improve coverage" into a measurable, automated, and demonstrably real result.
27 Sep 2026
