From Logs to Fixes: Teaching AI to Maintain Software
summary
Built a pipeline that freezes logs, feeds them to an LLM agent (Claude), maps trace lines to source files, generates fixes, creates GitLab issues and merge requests, and runs CI — all orchestrated with n8n orchestration. Demo uses a tiny Java blog on a €7 VPS; agent finds and corrects typos (wrong casing and a redirect key), produces a descriptive issue and MR, runs tests (16 passing) and deploys. Needs logs as source of truth, snapshots, test suites and guardrails; it struggles with frontend, architecture and very large codebases despite automatic MR generation.
AI-written summary · about
chapters (11)
- 0:00🚀 Build Stuff Conference – Get Your Tickets!
- 0:30Introduction: From Side Project to AI Maintenance Idea
- 3:40The Maintenance Problem in Software Teams
- 7:00Traditional Debugging vs AI-Assisted Flow
- 10:20Pipeline Overview: Logs → Fixes → PRs
- 14:00Demo Setup: Blog App with Intentional Bugs
- 18:30Running the AI Pipeline (n8n + LLM Agent)
- 23:00AI Detects Issues and Generates Fixes
- 27:00Reviewing AI-Generated Pull Requests
- 30:00What Works vs What Doesn’t (Limitations)
- 33:00Final Takeaways: AI as Assistive, Not Autonomous
related talks
19mHands-on Practice
35mFighting logs with Log Detective
38mEnhanced Log Analytics Using Generative AI
34mI Break Things, AI Fixes Them: Building a Self-Healing CI/CD Pipeline
46mLeveraging AI and Automation in DevOps
31mAgents at Work: How GitLab Is Redefining AI Adoption for Engineering Teams
37mHow AI helped us ship updates in a Linux distro
17mFighting AI with AI
