Ediky WorkflowHow the whole platform actually works.

Architecture, pipelines, APIs, and the decisions behind them — the single source of truth for every repository in the Ediky ecosystem.

Start with the Introduction

System Architecture

API Reference

🗺️

Ecosystem map

Four private repos — the website, ediky_tools, Docy, and these docs — and exactly how data flows between them, from an OA screenshot to a judged submission.

🏗️

Architecture, honestly

React 19 SPA on Cloudflare Pages, Firebase Auth + Firestore, self-hosted Judge0, and Workers for cron — with the tradeoffs stated, not hidden.

📦

Content at scale

Why the question bank ships as sharded JSON indexes + on-demand markdown bodies instead of import.meta.glob — and the 20,000-file limit that forced it.

🧪

OA Simulator internals

Server-anchored timers, one-way section locks, marking schemes, and the background Worker that finishes judging even after the tab closes.

🤖

The AI pipeline

One provider chain — Gemini → Groq → Cerebras → Mistral → NVIDIA → GitHub → Cohere → Ollama → OpenRouter — behind grading, tutoring, review, and these docs' own assistant.

🏭

The content factory

qgen for DSA (fuzzers, brute-force verifiers, adversarial cases), plus the CS, Aptitude, and SQL pipelines that validate and ship straight into the app.

🔐

Auth, data & rules

The Firebase authentication flow, the full Firestore schema, and the owner-only security rules that make "no backend" safe.

⚡

Performance playbook

Session-cached index shards, dependency-free row virtualization, deferred search, lazy routes — the ~300KB gzip the bundle lost, and how.

🤝

Contributing

Development setup, coding standards, the content linter that gates every deploy, and a troubleshooting guide for the failures we've already met.