ReleaseGuardian

Every production outage starts the same way. Someone deployed code without full context. The PR was in GitHub, the ticket was in Jira, the runbook in Confluence, past incidents buried in a folder nobody checks, and system health on a Grafana dashboard. Nobody reads all of it. The deploy decision gets made by one engineer who skimmed the PR, saw CI was green, and hit merge. 70% of production incidents are caused by changes. Teams then spend 30 to 60 minutes scrambling to gather the same context they should have reviewed before deploying. The data always existed - it just was never connected at decision time. ReleaseGuardian fixes this with a three-agent pipeline powered by Google Gemini. Agent 1 - Context Builder (Gemini Flash): Ingests raw artifacts - PR diffs, Jira tickets, service runbooks, past incident reports, and dashboard screenshots. Produces a structured release profile covering services touched, dependencies, affected user flows, risky code patterns, and similar past incidents. Agent 2 - Risk Analyst (Gemini Pro): Scores risk across six evidence-backed dimensions - code change scope, dependency risk, historical pattern match, test coverage, operational readiness, and current system health. Maps the blast radius showing which services, users, and teams are in the danger zone. Predicts the most likely failure modes with probabilities and severity. Agent 3 - Recovery Planner (Gemini Flash): Generates a complete recovery package - pre-deploy checklist, step-by-step rollback plan with trigger criteria and time estimates, monitoring thresholds you can paste into Grafana, and ready-to-send Slack messages, manager emails, and on-call handoff notes. Five inputs. Three agents. Five outputs. Under 60 seconds. No more disconnected context. No more scrambling after the fact. Built with React, FastAPI, and Google Gemini (Flash + Pro).