Pipeline Doctor

Pipeline Doctor auto-diagnoses GitLab CI failures with AI, suggests .gitlab-ci.yml fixes, auto-creates merge requests, finds similar past issues via vector search, and notifies devs instantly.

Comment 2

Pipeline Doctor – Your AI-Powered CI/CD Assistant 🩺🚀

🚀 Inspiration

As a software engineer, I frequently run into failed GitLab pipelines, often accompanied by cryptic and overwhelming logs. Pinpointing the root cause feels like searching for a needle in a haystack. Debugging becomes even more time-consuming when I have to rely on SREs for support.

I wanted to remove this friction and empower developers like myself to self-diagnose and fix CI/CD issues faster and more confidently. That’s how Pipeline Doctor was born—your friendly AI-powered companion that understands, analyses, and heals your pipelines.

🩺 What It Does

Pipeline Doctor is an AI-powered CI/CD assistant that:

  • 🔍 Analyses failing pipeline jobs using Gemini 2.0 Flash
  • 🧠 Performs similarity analysis on past failures using MongoDB Atlas Vector Search
  • 🛠️ Suggests AI-generated fixes to .gitlab-ci.yml (when config-related)
  • 🤖 Auto-creates a merge request with:
    • Root cause summary
    • Past similar job context
    • A fix diff patch for .gitlab-ci.yml
  • 📬 Sends this as a GitLab MR comment, or email (if not MR-linked)
  • 🏷️ Labels AI-created MRs with ai-generated and job ID reference

It not only saves developer time but also improves long-term pipeline health and debugging capabilities.

🛠 How I Built It

  • Tech Stack: Java 17 + Spring Boot
  • AI Analysis:
    • Used Gemini 2.0 Flash to summarize pipeline failures
    • Generated embeddings for log data
  • Vector Search:
    • Embedded historical job logs in MongoDB
    • Queried for top 3 similar failures using MongoDB Atlas Vector Search
  • Automation:
    • Used GitLab REST APIs to post comments and auto-raise MRs
  • Deployment:
    • Hosted on Google Cloud Run
    • Managed secrets via Google Secret Manager

⚠️ Challenges I Faced

  • First time working with MongoDB Atlas Vector Search—took time to learn the query structure
  • Choosing the best secret management tool took careful evaluation
  • Debugged a major MongoDB authentication error before the demo deadline

🏆 Accomplishments I'm Proud Of

  • 🎯 First solo project combining AI + Vector DB + CI/CD + Cloud
  • 🔧 Created an end-to-end tool that is immediately usable in real developer workflows
  • 🧠 Took a personal pain point and solved it using modern AI tooling
  • 🚀 Got hands-on experience with full-stack DevOps & ML engineering

📚 What I Learned

  • 🧾 Prompt engineering for Gemini 2.0
  • 🔍 Vector similarity search with MongoDB Atlas
  • 🛠️ GitLab APIs for CI/CD automation
  • ☁️ Google Cloud Run and Secret Manager for production deployment

🌱 What's Next for Pipeline Doctor

  • ☁️ Build a cloud-hosted version to increase adoption, as the current version is self-hosted, and the developers need to use their infrastructure to use the application as a CI/CD component
  • 🔘 Add user toggle controls for auto-analysis and auto-fix generation
  • 🔄 Extend support to other CI/CD platforms
  • 🧠 Use RAG (Retrieval-Augmented Generation) to make AI fixes even smarter

Built With

+ 1 more
Share this project:

Updates