ChangeBlast

Problem A small code or database change can have a much larger impact than expected. Developers and reviewers often have to manually trace dependencies across database schemas, source code, APIs, tests, documentation, and existing data. This becomes especially dangerous for destructive database changes, where a migration can execute successfully while silently breaking application behavior or destroying data. Solution ChangeBlast is an AI-powered pre-production change-impact analyzer designed to answer one question before deployment: “What is the blast radius of this change?” Given a proposed change, ChangeBlast analyzes the project artifacts and traces its dependencies across multiple layers. It identifies affected database objects, source-code references, API endpoints, tests, documentation, and potential data-loss scenarios. The findings are presented as an evidence-backed impact report with a visual impact map, risk assessment, validation results, and recovery guidance. For our demonstration, we analyze a seemingly simple migration: ALTER TABLE users DROP COLUMN phone; ChangeBlast discovers that the phone column is referenced across the ShopLite application. It identifies four affected API endpoints and validates the predicted impact in an isolated database environment. The original test suite contains 26 passing tests; after applying the destructive migration in the temporary environment, only 4 remain passing and 22 fail because the expected column no longer exists. ChangeBlast also detects the data-loss consequence. Existing phone values are removed from the migrated database, and simply adding the column back does not restore the original values. The system therefore distinguishes between reversing a schema operation and actually recovering lost data, recommending restoration from a verified backup or authoritative source. IBM Bob 2.0 IBM Bob 2.0 is central to our development and analysis workflow. Its agentic capabilities allow us to work across project artifacts, trace dependencies, reason about the proposed change, and generate the analysis. We then use controlled validation to verify the predicted consequences rather than relying solely on AI-generated conclusions. Target Users & Value ChangeBlast is designed for software developers, code reviewers, database engineers, and engineering teams working with evolving applications. It can help surface hidden dependencies before deployment, provide evidence during code review, identify destructive changes, and make recovery requirements explicit. The MVP demonstrates the core workflow using Python, Flask, SQLite, pytest, and IBM Bob 2.0. Future versions could integrate with Git repositories and CI/CD pipelines and support additional languages and database systems. ChangeBlast: see the blast radius of your change before production does.