Problem: Most backend performance issues in production stem from poorly implemented pagination — unbounded queries, deep offsets, missing limits, and unstable sorting — that go undetected until they cause real scaling failures.

Solution: Built Cloud Meter, a production-grade CLI tool published to npm that statically analyzes backend codebases (no runtime required) using AST parsing to detect pagination anti-patterns, score the codebase from 0–100, and deliver actionable fix recommendations.

Key Features:

☁️ Static AST Analysis: Scans JS/TS files, SQL queries, and ORM patterns without executing any code
📊 Scoring Engine: 0–100 score with severity-based deductions — Missing LIMIT (−30), Deep OFFSET (−20), No ORDER BY (−15), and more
🔍 12 Detection Rules: Catches missing LIMIT, deep OFFSET, unstable cursors, unbounded endpoints, dynamic sort injection, infinite scroll risks, and concurrent write exposure
🛠️ Self-Guiding CLI: Every command outputs what to run next — fully interactive or headless CI mode
⚙️ Config System: Project-level cloud-meter.config.json with saved defaults for path, output mode, severity threshold, and ignored directories
🚦 CI/CD Integration: --mode ci exits with code 1 when findings exceed configured severity threshold — drop into any pipeline

Technical Implementation:

Built CLI with Node.js and TypeScript, published as cloud-meter on npm with global install support
Implemented AST-based pattern matching for Prisma, Mongoose, Sequelize, TypeORM, raw SQL, and GraphQL Relay-style resolvers
Designed a modular rulebook with configurable severity weights and heuristic penalties for dynamic sorting and concurrency risk
Added framework detection for Express, NestJS, Fastify, and Koa route/controller/service layers
Supported monorepo structures with auto-detection of apps/api, services/api, and backend directories
Persisted analysis cache for recommend command to serve fix recipes from the last scan without re-running

Impact: Delivered a zero-runtime, production-grade CLI that catches pagination scaling risks before they hit production — installable in one command, CI-ready out of the box, and covering the full ORM, SQL, and GraphQL ecosystem.

PORTFOLIOCOOKING.