Develop Greenly

Develop Greenly is a software sustainability observatory for CI/CD workflows. It collects metrics from GitHub Actions, estimates energy consumption and CO2 emissions, stores the data in SQLite, and presents repository-level insights in an interactive dashboard. CO2 emission is estimated using simplified version of Software Carbon Intensity (SCI) specification by Green Software Foundation (GSF):

CO2 = workflow duration x estimated runner energy use x carbon intensity

The dashboard is hosted on AWS EC2, available live at http://63.179.89.6/.

Features

• Collects CI/CD metrics from GitHub Actions through a backend API.
• Stores workflow history in SQLite for persistent dashboard data.
• Tracks repository, branch, commit, runner type, duration, energy, CO2, timestamp, metric type, step name, and workflow phase.
• Compares build and runtime smoke-test cost across repositories.
• Groups related metrics by repository and commit.
• Shows total CO2, total energy, average duration, tracked repositories, highest-cost run, and latest received run.
• Visualizes CO2 trends across recent commits.
• Compares repository emissions using charts and summary cards.
• Supports step-level CI breakdowns for phases such as dependencies, build, tests, runtime, deploy, and setup.
• Provides optimization recommendations based on workflow patterns, step-level hotspots, missing runtime metrics, missing step data, frequent runs, and CI duration regressions.
• Runs locally or on AWS EC2 using Docker Compose.

Tech Stack

• Frontend: React, Vite, Recharts, CSS
• Backend: Node.js, Express
• Database: SQLite with better-sqlite3
• CI Source: GitHub Actions
• Deployment: Docker Compose on AWS EC2

Architecture

GitHub Actions workflow
        |
        | POST /api/green-metrics
        v
Node.js / Express backend
        |
        | stores metrics
        v
SQLite database
        |
        | GET /api/green-metrics
        v
React dashboard

Dashboard Views

The dashboard includes:

• KPI summary for total CO2, energy, average duration, and tracked repositories.
• Trend chart showing CO2 across recent commits.
• Repository comparison showing cumulative impact per repository.
• Optimization recommendations with severity, evidence, and suggested action.
• Step-level CI breakdown showing where workflow energy is spent.
• Recent workflow runs table combining build, runtime, and total metrics by commit.

Getting Started Locally

Clone the repository:

git clone https://github.com/malihanawshin/Develop_Greenly.git
cd Develop_Greenly

Create an environment file for the backend token:

METRICS_API_TOKEN=your-shared-token

Start the full app with Docker Compose:

docker-compose up -d --build

Running Without Docker

Start the backend:

cd green-dev-backend
npm install
METRICS_API_TOKEN=your-shared-token npm start

Start the frontend:

cd green-dev-dashboard
npm install
npm run dev

GitHub Actions Integration

Each repository that sends metrics should define these secrets:

GREEN_METRICS_URL=http://your-server/api/green-metrics
GREEN_METRICS_TOKEN=your-shared-token

The workflow sends JSON metrics to the backend:

{
  "repo": "owner/repository",
  "branch": "main",
  "commit": "commit-sha",
  "duration": 42,
  "runner_type": "Linux",
  "metric_type": "ci_build"
}

The backend estimates energy and CO2, adds a timestamp, and stores the row in SQLite.

Example response:

[
  {
    "repo": "malihanawshin/LAN-Chat",
    "branch": "main",
    "commit": "be7787a42f9561fad8c17cfdbb4e81ac461d36c1",
    "duration": 15,
    "runner_type": "Linux",
    "metric_type": "ci_step",
    "step_name": "compile_server",
    "phase": "build",
    "energy": 0.0125,
    "co2": 5.94,
    "timestamp": "2026-07-01T18:08:37.987Z"
  }
]

Deployment

The app is deployed on one AWS EC2 instance with Docker Compose.

On the EC2 instance:

git clone https://github.com/malihanawshin/Develop_Greenly.git
cd Develop_Greenly
docker-compose up -d --build

After pushing changes:

cd ~/Develop_Greenly
git pull origin main
docker-compose up -d --build

The SQLite database is stored in a Docker volume so dashboard data survives container rebuilds.

Upcoming Features

• Region-aware carbon intensity using live or configurable grid data.
• Improved SCI calculation with CPU, memory, and runner utilization inputs.
• AI-assisted recommendation summaries based on rule findings and historical trends.
• Repository onboarding templates for Node.js, Python, Java/Maven, C++, Next.js, React Native, and mobile projects.
• Exportable CSV/JSON reports for applications, portfolio documentation, and sustainability analysis.
• Advanced filters by repository, branch, date range, metric type, phase, and runner type.
• Regression alerts when CI energy or duration increases significantly.
• Saved baselines for comparing commits, branches, and repositories over time.
• Authentication and dashboard access control for hosted deployments.

Current Limitations

• Energy and CO2 values are estimates, not direct hardware measurements.
• The model uses a fixed runner power assumption and fixed carbon factor.
• CPU load, memory usage, runner region, and grid carbon intensity are not yet included.
• GitHub-hosted runner data is represented by runner operating system only.
• Recommendations are rule-based and explainable, not generated by an external AI model.

License

This project is under the MIT License.