
Developers responding to a failure often search logs, trace source code, inspect dependencies, and run tests in separate steps. That fragmentation makes incidents slow to resolve and makes it easy to accept a plausible fix without evidence or a regression test. DevGuard brings those steps into one reviewable workflow for software engineers and small development teams. The user starts with a bug report, log, and repository. IBM Bob IDE is the central workspace for focused investigations: analyzing the stack trace, tracing the relevant code path and data flow, and checking dependencies and existing tests. The findings are combined into ranked hypotheses with repository-relative file and line references. The workflow then calls for a small fix, a regression test that reproduces the original failure, and validation with the project test suites. A companion Python CLI gathers bounded, read-only evidence and writes a reviewable report without editing the target repository. A local dashboard can display DevGuard JSON reports and scan an uploaded source or log file in the browser. The sample incident is an empty order that reached average_price([]) and caused division by zero; the included fix defines the empty case as 0.0 and has a regression test. Local and GitHub Actions monitoring can run the tests again to catch a recurrence. DevGuard is distinctive in connecting diagnosis, cited evidence, code review, regression testing, and monitoring. The evidence collector limits file and log sizes, excludes common secret paths, redacts common secret assignments, and checks proposed diffs against current source. The sample uses synthetic data.
27 Sep 2026