Inspiration

CI/CD failures often slow down development more than they help. We found ourselves repeatedly going through long logs to debug simple issues like missing dependencies or misconfigurations. This repetitive and time-consuming process inspired us to build a solution that could automate debugging and learn from past mistakes.

What it does

Our AI DevOps Failure Memory Agent analyzes CI/CD pipeline logs, identifies the root cause of failures, and suggests fixes instantly. It also stores past errors and retrieves similar issues, helping us debug faster over time.

How we built it

We built a full-stack system:

FastAPI for the backend API Streamlit for the user interface A log parser to extract error and context Sentence Transformers + FAISS for storing and searching past errors A rule-based AI engine with optional LLM (Ollama) for generating fixes

The system processes logs, searches memory, analyzes errors, and returns actionable suggestions.

Challenges we ran into

Parsing unstructured and inconsistent CI/CD logs Ensuring accurate similarity matching in the vector database Avoiding incorrect or vague fix suggestions Balancing speed with memory-based search Integrating multiple components smoothly

Accomplishments that we're proud of

Built a working AI-powered debugging assistant end-to-end Successfully implemented memory-based learning using FAISS Achieved real-time error analysis with useful fix suggestions Designed a scalable and modular architecture Created a tool that can be extended to real CI/CD environments

What we learned

How to build AI systems that improve using past data Practical use of vector databases and semantic search Handling real-world DevOps logs and failure patterns Integrating backend, frontend, and AI components Importance of performance and reliability in developer tools

What's next for AI_DevOps_Failure_MemoryAgent

Automatic fixing of CI/CD errors Integration with GitLab/GitHub pipelines Slack/Email alerts for failures Multi-project and team-based memory sharing Advanced LLM-based analysis for deeper insights

Built With

  • ai-powered
  • api
  • communication
  • database
  • faiss
  • fastapi
  • git
  • gitlab
  • llm
  • ollama
  • project
  • python
  • requests
  • streamlit
  • transformers
  • ui
  • vector
Share this project:

Updates