• 

  In this image Rosetta has successfully fixed the error in the code and pushed it to the GitLab repo.

• 

  In this image Rosetta is waiting for the user to hit "Fetch Live GitLab Log" to get the crash details from the repo.

Project Story

Inspiration

Every developer knows the frustrating feeling of pushing code to the cloud, waiting for the CI/CD pipeline to build, and watching it flash a bright red Failed badge. Instead of working on new features, your entire momentum grinds to a halt. You are forced to dig through thousands of lines of messy terminal logs just to find a single missing variable or an empty array exception.

I built Rosetta Agent to completely automate this tedious loop. I wanted to create an intelligent tool that connects cloud logs directly to automated code fixes, eliminating manual debugging downtime.

What it does

Rosetta Agent is an autonomous, self-healing DevOps assistant. When a cloud deployment pipeline fails, Rosetta automatically:

1. Intercepts the cloud environment to fetch the exact crash logs.
2. Diagnoses the root cause of the failure using generative AI.
3. Generates a precise, targeted Python code patch.
4. Pushes the fix directly back to the GitLab repository via API.
5. Monitors the live container build in real-time until the project successfully passes.

How we built it

The core engine is built entirely in Python and structured into a clean automation framework.

System Architecture Breakdown

The Automated Workflow Loop

# A conceptual look at how Rosetta automates the repair loop
def rosetta_pipeline_heal(project_id):
    raw_logs = gitlab_api.fetch_failed_logs(project_id)
    clean_error = log_parser.isolate_traceback(raw_logs)
    suggested_patch = gemini_client.generate_code_fix(clean_error)
    gitlab_api.apply_patch_and_push(project_id, suggested_patch)
    return "Pipeline Monitoring Active..."

### Accomplishments that we're proud of
Advanced System Prompting: Gained deep experience in setting guardrails and structural processing logic to guarantee reliable, deterministic AI code generation.

DevOps Workflows: Learned how remote cloud runners, container build lifecycles, and RESTful Git APIs interact under the hood during a pipeline lifecycle.

### What's next for Rosetta Agent
[ ] Multi-Platform Integration: Expanding API connectors to natively support GitHub Actions, CircleCI, and AWS CodePipeline.

[ ] Multi-File Context Aggregation: Scaling cognitive scopes so Rosetta can analyze system-wide cross-file dependencies to resolve multi-file architectural bugs.

[ ] Production Cloud Deployment: Transitioning the engine from localhost to a permanent cloud native host (such as Google Cloud Run) for public web accessibility.

Built With

• gitlab
• googlegenaisdk
• python
• streamlit
• yaml