
Bug2PR is an autonomous bug-to-PR pipeline that eliminates the 5-hour gap between bug discovery and a merged pull request. Built on IBM Bob, it uses 8 specialist AI agents working in sequence - log parsing, file localization, code retrieval, root cause analysis, fix generation, test writing, security auditing, and PR creation - to deliver a complete, tested GitHub pull request in roughly 60 seconds. The pipeline is orchestrated by LangGraph as a state machine, with IBM Bob driving the three most reasoning-intensive tasks: root cause analysis, regression test generation, and security auditing. Faster, cheaper models like Groq handle lighter tasks such as file scanning. A four-layer fallback chain (Bob → Gemini → Groq → Ollama → rule-based) ensures the pipeline never breaks. Bug2PR is both reactive and proactive. Developers can paste a stack trace into the War Room - a real-time dashboard that streams each agent's status via WebSocket - or connect it to a GitHub repository for proactive scanning. The Risk Report feature analyzes all code in a project, ranks issues by severity, and provides a "Fix this" button that runs the full pipeline on any finding with zero manual setup. Every run is tracked on the MTTR (Mean Time To PR) Dashboard, showing a 99.6% reduction from the 5-hour manual baseline. Slack notifications close the loop by alerting the engineering team the moment a PR is ready. Built in 9 documented sessions with IBM Bob, Bug2PR proves that autonomous debugging isn't just possible - it's production-ready.
17 May 2026