THE PROBLEM Every developer has broken something they thought was a small, isolated change. A single rename, a type change, or a schema tweak can silently ripple through dozens of files, API contracts, database tables, and test suites — and the developer usually only discovers the blast radius after something breaks in production, or worse, after a customer notices. Existing tools review code after it's written; almost nothing helps a developer understand the consequences of a change before a single line is typed. THE SOLUTION ChangeBlast answers one question before you touch any code: "What will this change actually affect?" A developer describes a proposed change in plain English — for example, "Replace User.id from Integer to UUID" — and ChangeBlast statically scans the target repository and returns a complete, evidence-backed picture of the impact: every directly and indirectly affected file, every touched API contract, every database table or migration required, every test that needs updating, and a deterministic risk score explaining exactly why the change is risky. It then produces an ordered, phased change plan so the developer knows what to do first. HOW IT WORKS The backend runs a repository-analysis pipeline: it parses source files, API route definitions, database models/migrations, and test suites in a seeded sample codebase, and classifies every finding into evidence objects (file, line number, relationship, description) — never an unproven claim. These are grouped into impact categories (code, API, database, frontend, tests, history), fed into a deterministic risk engine (fixed point values for public API exposure, database migrations, auth-path changes, dependent-file count, and test coverage), and finally turned into an ordered, human-readable change plan. Everything is exposed through a versioned REST API (POST /api/v1/analysis, GET /api/v1/analysis/{id}, GET /api/v1/analysis/{id}/status) so the frontend — or any other client — never has to guess at backend internals. The frontend is a React application that walks the developer through the full flow: entering the change, watching analysis progress, and landing on a dashboard with an interactive SVG blast-radius graph, a risk panel, a tabbed impact explorer, an evidence browser, and the recommended change plan — all rendered directly from the backend's real JSON response, with zero hardcoded demo data. TARGET USERS Any developer or team working in a codebase large enough that "what does this touch?" is no longer obvious — particularly useful before refactors, schema migrations, ID/type changes, or any change that crosses the frontend/backend/database boundary. WHAT MAKES IT DIFFERENT Most AI coding tools review or write code. ChangeBlast is deliberately not a chatbot and not a code-review tool — it is a single, focused pre-implementation workflow. Its risk score is deterministic, not an LLM guess, so it's reproducible and explainable. Every finding is required to carry real evidence (file + line + description) rather than a generic AI claim, which makes the tool trustworthy rather than merely plausible. The system was also built to be effective and efficient for a hackathon timeline: no external database, no heavy multi-agent orchestration, and a clean API contract that let backend and frontend be built independently by two people in parallel and integrated in minutes.
Category tags: