ReplayGuard

ReplayGuard — An AI Remediation Laboratory for Runtime Contract Drift When an external API silently renames a response field — `status` to `paymentStatus` — the code still compiles. Unit tests stay green because they use stale mocks. Documentation still shows the old contract. Then production crashes because a downstream service reads a field that no longer exists. This is contract drift, one of the costliest and least-visible failure modes in modern microservices. Human code reviewers miss it. Static analyzers don't understand runtime behavior. Existing AI coding assistants generate a single patch and ask the developer to trust it. ReplayGuard takes a different approach. It is a decision-support workspace built on top of IBM Bob 2.0's repository analysis. When pointed at a repository, ReplayGuard turns Bob's findings into an interactive laboratory that lets a developer see the failure, understand the blast radius, compare remediation strategies, and verify the selected fix before merging. Who it is for: Software engineering teams working on microservices, API-driven systems, or any codebase where breaking changes can propagate silently across service boundaries. The primary user is a developer or tech lead reviewing a pull request or investigating a production incident. How they interact with it: A developer opens ReplayGuard and immediately sees the incident header, the affected workspace, and the exact contract change. They click "Simulate Cascade" to watch the failure propagate visually from the origin service to the blocked downstream consumer. They click "Replay Failure" to step through the runtime failure sequence, frame by frame, with file-and-line citations for every step. They inspect the Documentation Drift panel, which shows the stale API specification side-by-side with the correct contract. They inspect the False-Green Test Trap panel, which explains why the existing test suite passed while the real contract had changed. Then the developer reaches the Multi-Hypothesis Remediation Lab. Instead of one fix, IBM Bob 2.0 explored three competing architectural strategies in parallel — a Compatibility Adapter (backward-compatible, leaves technical debt), a Strict Contract Migration (clean, breaks unmigrated consumers), and a Contract Boundary Guard (runtime schema validation). Each card shows files affected, lines changed, risk level, legacy support, and trade-offs. Clicking between them dynamically updates a syntax-highlighted diff viewer and a consequence preview. The developer selects a strategy and clicks Apply This Remediation. The verification gate runs the test suite and shows the transition from FAIL to PASS. The proof is real: the repo contains a runnable Vitest suite where one test fails against the broken code and passes after the fix. What makes it unique: Most AI coding tools give you a fix. ReplayGuard gives you three verified choices with evidence. It is not an auto-fixer. It is a decision-support layer that respects the engineering judgment of the human who will live with the consequences of the choice. Impact: It reduces change-risk review from an estimated 45 minutes of manual dependency tracing to roughly 30 seconds of evidence-backed decision-making.