Agent Anansi

• 

  Detailed Architecture

• 

  gitlab ci actions(all 6 agents)

• 

  Simple Architecture

Inspiration

As someone deeply passionate about DevOps and AI, I was frustrated by the fragmented and reactive nature of traditional CI/CD workflows. While automation is widespread, intelligence is often lacking — developers still manually review code, hunt down security issues, or burn out in silence. Inspired by African mythology, I envisioned Agent Anansi as a digital spider weaving intelligent threads through the CI/CD pipeline. I wanted to create a practical, scalable, and human-centered DevOps assistant powered by AI agents that could do more than just automate — it could think, optimize, and care.

What it does

Agent Anansi is a fully automated, AI-powered CI/CD assistant that integrates directly into GitLab pipelines. It consists of 6 specialized agents, each handling a different facet of modern software delivery:

1. 🔍 Issue Triage Agent – Automatically categorizes and prioritizes issues.
2. 🔀 Code Review Agent – Reviews merge requests and code changes.
3. 🔐 Security Scan Agent – Scans the codebase for vulnerabilities and insecure dependencies.
4. ⚡ CI Optimizer Agent – Analyzes pipeline performance and recommends optimizations.
5. 📚 Documentation Generator Agent – Auto-generates API and project documentation.
6. 💚 Burnout Detector Agent – Monitors developer patterns to detect signs of burnout.

Each agent generates structured JSON reports, and all insights are aggregated into a single job that visualizes the results. This creates an intelligent feedback loop on every commit or merge request.

How I built it

The project is built with:

• Python 3.11 for agent logic and REST endpoints.
• Google Cloud (Google AI Studio, Google Agent Development Kit + Cloud Run) to deploy the agents and host the intelligence layer.
• GitLab CI/CD for deep integration and automation across code analysis and deployment.
• Google Gemini Flash 2.0 as the core LLM for natural language understanding and reasoning.

I wrote modular HTTP handlers for each AI agent and designed the pipeline to be fail-tolerant and trigger-based (main branch, MRs, or manually). All AI responses are formatted into machine-readable reports stored as CI artifacts.

Challenges I ran into

• Burnout Detection Design: Quantifying developer wellness using commit metadata was novel, and required creative metrics design to avoid false positives.
• YAML Management: Handling shared configuration for multiple agents and ensuring secure, minimal Docker environments without bloating image sizes was tricky.

Accomplishments that I'm proud of

• Created 6 fully working AI agents with real endpoints and live results integrated into GitLab pipelines.
• Designed a developer wellness AI agent — a feature I haven’t seen in any mainstream CI/CD tool.
• Successfully deployed everything to Google Cloud Run, making the platform truly plug-and-play for any GitLab team.
• The final demo script brings the results to life with detailed feedback, artifact reports, and a beautiful console showcase.

What I learned

• The power of multi-agent AI systems is in their composition. Each agent is simple, but together they orchestrate complex DevOps decisions.
• Even solo, you can ship a full-stack AI product with CI/CD, cloud infra, and usable interfaces if you focus on automation and reuse.
• Monitoring developer wellness can and should be part of the CI/CD process. It’s not just about code — it’s about people.

What's next for Agent Anansi

• Natural Language Feedback: Add conversational UIs for developers to query the agents' findings via comments or Slack.
• More Metrics: Expand burnout detection with deeper git analytics and optional developer journaling.
• Team Dashboards: Build a web-based dashboard showing trends across pipelines, team health, and optimization over time.

Built With

• cloudrun
• gemini-2.0-flash
• gitalb-ci
• google-agent-development-kit
• python