DevFlow AI

Problem: Software maintenance forces developers through a slow, sequential manual workflow: read code, identify bugs, review tests, check security, fix issues, write tests, run tests, investigate failures, repeat. Each step is manageable alone, but together they consume hours and produce inconsistent results. No single tool manages the complete review-debug-test-verify cycle, and manual workflows leave no measurable baseline for improvement. Solution: DevFlow AI automates the entire developer maintenance workflow using a two-tier AI agent architecture built on IBM Bob 2.0. A main orchestrator agent manages workflow state and delegates tasks to four specialized subagents running in parallel: a Code Review Agent, a Security Agent, a Test Analysis Agent, and a Documentation Agent. Their findings are automatically aggregated, deduplicated, and prioritized. The platform then generates a fix plan, presents it to the developer for approval, applies safe fixes with automatic backup and syntax verification, generates new test cases, runs the test suite, analyzes failures, and produces a structured final report. Target Users: Software developers and engineering teams who regularly perform code review, security auditing, and maintenance on Python-based repositories. How It Works: The developer points DevFlow AI at a repository. The orchestrator inspects the codebase, dispatches all four analysis agents, collects their findings, and guides the developer through a minimal approval step before applying any code changes. The entire pipeline from repository scan to final report runs with at most 4 developer interactions. What Makes It Unique: DevFlow AI is the only tool that connects the full cycle — analysis, fix planning, human approval, code modification, test generation, test execution, and final reporting — in a single automated pipeline. It detects 85% of seeded issues (11 of 13), generates 14 test cases automatically, applies fixes with zero regression risk (syntax-checked + auto-reverted on failure), and produces a final Markdown report with before/after metrics. All 13 acceptance criteria for the IBM Developer Workflow Automation Challenge are verified.