Bob Change Impact Mode

Today, developers change existing code by exploring it manually — a slow, error-prone process where dependencies get missed and bugs only surface after the change is already merged. Bob Change Impact Mode turns that guesswork into a guided, auditable workflow: Analyze → Execute → Validate. When a developer describes a change, three specialist agents run in parallel: • Dependency Agent — walks the repository and identifies every file and module affected by the change • Database Agent — inspects schema and migration files to flag impacted tables, each tagged with a risk level (HIGH/MEDIUM/LOW) • Test Agent — cross-references affected files against the test suite to surface exactly which tests need to re-run Their outputs are merged into a single risk summary and a numbered implementation plan. Critically, nothing is executed without human approval — the developer reviews the plan before Bob touches any code. Once approved, Bob executes the change, runs the tests, retries once on failure, and generates a final report combining a git-diff-style summary with pass/fail test results. The entire pipeline is orchestrated with LangGraph as a 7-stage state machine, exposed through a FastAPI backend that streams every stage live to a single-page frontend via Server-Sent Events — no mock data, no simulated steps. This project was built entirely inside IBM Bob 2.0, using Agent mode to generate the LangGraph agents, wire the backend, and build the live frontend from plain-English specifications — a real demonstration of Bob's ability to take a project from concept to a working, demo-ready application.