ReleaseGuard

ReleaseGuard is an AI-assisted release readiness platform designed to help software teams evaluate whether a release is ready for deployment before it reaches production. Software releases often require several checks across different areas, including code changes, dependencies, environment configuration, database migrations, test coverage, and deployment preparation. When these checks are performed manually or across disconnected tools, important risks can be missed and the release process becomes harder to understand and repeat. ReleaseGuard brings these checks into one workflow. A user uploads a release bundle as a ZIP file containing release metadata, a diff patch, package information, and source files. The platform validates the bundle and creates a unique analysis run. The release is then processed through eight deterministic analysis stages: 1. Ingest Bundle 2. Diff Analysis 3. Dependency Check 4. Environment and Configuration Check 5. Database Migration Check 6. Test Coverage Gap Analysis 7. Deployment Checklist 8. Report Generation The application streams the progress of these stages to the browser using Server-Sent Events, allowing the user to see each analysis step as it completes instead of waiting for an unexplained final result. After the analysis finishes, ReleaseGuard aggregates the findings into a consolidated release report. The report displays a readiness score, overall severity, individual warnings, detected risks, and recommendations for addressing those risks. IBM watsonx.ai is integrated into the final report-generation stage. IBM Granite is used to generate human-readable release notes from the verified results produced by the deterministic analysis pipeline. The underlying release metrics and readiness score come from the analysis pipeline rather than being invented by the model. If the AI service is unavailable, the application can still complete the analysis using its fallback behavior. ReleaseGuard also provides a Past Reports view so users can open completed analyses without running the release pipeline again. Reports are persisted as JSON snapshots, making completed release results easy to retrieve during the prototype workflow. The target users are software developers, release engineers, QA teams, and engineering teams that need a clear pre-deployment view of release risks. The project focuses on making release readiness more transparent: instead of simply producing a final score, ReleaseGuard shows how that result was produced, which areas contain risks, and what should be reviewed before deployment. The prototype is implemented with React and Vite on the frontend, Node.js and Express on the backend, Server-Sent Events for live pipeline updates, JSON-based persistence for reports, and IBM watsonx.ai for grounded release-note generation. The result is a single workflow for uploading a release, observing its analysis in real time, reviewing actionable findings, and producing concise AI-assisted release notes.