CodeZero Green AI

CodeZero is a VS Code extension that scans source files for energy-heavy coding patterns and suggests greener alternatives before they turn into production waste.

What this MVP does

• Scans the active editor with AST-based static analysis for JavaScript and TypeScript.
• Falls back to heuristic scanning for Python and Java when AST parsing is not available.
• Uses CO2.js to estimate the carbon impact of the file plus modeled waste from flagged patterns.
• Adds inline diagnostics, quick fixes, a CodeZero activity bar UI, a findings tree, and a full scan report panel.

Architecture

• Carbon library: CO2.js by The Green Web Foundation using the Sustainable Web Design model (swd, version 4).
• Static analysis: Babel AST parsing and traversal for JS/TS files, with heuristic fallback for non-AST languages in this MVP.

Included rules

• Polling timer: flags setInterval(..., 1000) and faster intervals.
• Nested loop: warns on nested loops that often scale poorly.
• Wildcard import: flags import * as ... because it tends to increase bundle size.
• Repeated array scan: flags .filter(...).length and recommends .some(...).
• Verbose logging: flags frequent console.log usage in hot paths.
• Remote call in loop: flags likely network or database calls inside loops.

These are heuristics, not full program analysis. The goal is to surface likely waste early and educate the developer while they code.

Getting started

npm.cmd install
npm.cmd run build

Then open this folder in VS Code and run the extension with F5.

UI

• Activity Bar: open the CodeZero icon to view the persistent sidebar.
• Dashboard: shows eco score, estimated CO2e, source size, transfer model, and engine details.
• Findings: clickable list of sustainability issues that jumps to the matching line in the editor.
• Status Bar: shows the current file's eco score.
• Command Palette: run CodeZero: Scan Current File for a full report webview.
• AI Refactor: run CodeZero: AI Refactor Current File to request a greener rewrite from the configured CodeZero hosted AI backend. Development fallback providers are still available through CodeZero: Configure AI Refactor, but published builds should keep provider API keys on a server, not inside the extension.

Hosted AI for Marketplace Builds

CodeZero defaults to the codezero AI provider, which calls codezero.hostedAiEndpoint instead of asking each user for an API key. Deploy the proxy from this repository's server/refactor-proxy.js behind HTTPS, set GEMINI_API_KEY in the hosting environment, then point codezero.hostedAiEndpoint at that URL before publishing.

Do not embed Gemini or OpenAI keys in the extension package. A published .vsix can be inspected by users, so shipped keys should be treated as public.

Demo Files

• demo/codezero-test.ts: intentionally inefficient sample to trigger findings.
• demo/codezero-green.ts: greener version of the same workflow for side-by-side comparison.

Example quick fixes

• Replace items.filter(predicate).length > 0 with items.some(predicate)
• Replace import * as utils from "./utils" with a named import placeholder
• Replace setInterval(..., 500) with setTimeout(..., 5000) guidance

Next steps

• Add ESLint custom rules as an optional export path for CI enforcement.
• Expand AST support to Python and Java with dedicated parsers.
• Connect the dashboard to an AI explanation service for deeper refactoring help.