Jan 30, 2026

Kord

A lightweight versioned content and data storage platform built on Nuxt 3 and Cloudflare edge ecosystem. Manages subtitles, prompts, notes, and structured JSON data with full version history.

NuxtNitroCloudflareEdge ComputingDrizzle ORM

Kord: Versioned Content & Data Storage on Cloudflare Edge

A lightweight content and data storage platform focused on receiving and managing subtitles, prompts, notes, and structured data, with complete version history.

🌐 Live Demo: https://kord.sedationh.top

Core Scenarios

1. Video Subtitle Archival

Save subtitle files with Canonical Keys to ensure one video only has one copy.

Input https://youtube.com/watch?v=dQw4w9WgXcQ → automatically parsed to youtube:dQw4w9WgXcQ
Next time you save subtitles for the same video, it updates the same resource instead of creating duplicates

2. Text Version Evolution

Perfect for content that requires iterative refinement:

Prompt tuning - Each modification creates an immutable snapshot
Translation polishing - Roll back to any previous version anytime
Draft auto-save - Real-time sync while editing

3. Structured Data Storage

Through the data: type, store JSON data for various lightweight storage scenarios:

Key	Use Case	Example Data
`data:blog-stats`	Blog likes count	`{"likes": {"post-1": 42}}`
`data:site-config`	Site configuration	`{"theme": "dark"}`
`data:reading-list`	Reading progress	`[{"title": "Book", "progress": 0.5}]`

4. Workflow Integration

Open API via Personal Access Token (PAT) for AI Agents or automation workflows - serving as a versioned storage backend for your automation pipelines.

Architecture Design

Metadata-Blob Separation Architecture - distributing data across different Cloudflare components based on access frequency.

┌───────────────────────────────────────────────────────────┐
│                     Cloudflare Edge                       │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐       │
│  │   D1 (SQL)  │  │  KV (K/V)   │  │  R2 (Blob)  │       │
│  │   Metadata  │  │   Drafts    │  │  Snapshots  │       │
│  │   Warm      │  │   Hot       │  │   Cold      │       │
│  └──────┬──────┘  └──────┬──────┘  └──────┬──────┘       │
│         │                │                │               │
│         └────────────────┼────────────────┘               │
│                          │                                │
│                ┌─────────┴─────────┐                      │
│                │  Nuxt 3 (Nitro)   │                      │
│                │   Edge Runtime    │                      │
│                └─────────┬─────────┘                      │
└──────────────────────────┼────────────────────────────────┘
                           │
                  ┌────────┴────────┐
                  │     Client      │
                  │   (Vue 3 SPA)   │
                  └─────────────────┘

Data Tiering

Tier	Storage	Purpose	Characteristics
Hot 🔥	KV	Current state	Real-time sync on every change
Warm 📊	D1	Metadata & summaries	Fast list page queries
Cold ❄️	R2	Immutable snapshots	Frozen on commit

Tech Stack

Layer	Technology	Description
Full-stack Framework	Nuxt 3	Vue 3 based full-stack framework
Server Engine	Nitro	Native Cloudflare Workers support, cold start <10ms
Metadata Layer	Cloudflare D1	Edge SQLite for indexes, versions, tokens
Active Content Layer	Cloudflare KV	Low-latency key-value storage
Archive Layer	Cloudflare R2	Object storage for immutable snapshots
ORM	Drizzle ORM	Edge-optimized, type-safe, zero runtime overhead

Canonical Key System

All user inputs are normalized into canonical keys for deduplication:

Input Type	Canonical Key	Content Format
`data:KEY`	`data:blog-stats`	JSON
YouTube URL	`youtube:VIDEO_ID`	Markdown
Bilibili URL	`bilibili:BVID`	Markdown
Other URL	`url:SHA256_HASH`	Markdown
Custom note	`note:SLUG`	Markdown

The same canonical key can only have one resource per user - achieving automatic deduplication.

Authentication

Dual-track authentication system balancing UI interaction and API automation:

1. UI Interaction (GitHub OAuth)

Sealed Session Cookie with edge-local decryption
No database query needed for verification

2. API Access (PAT)

For AI Agents, CLI, and automation scripts
Backend stores only SHA-256 hash, supports instant revocation

Key Features

✅ Content & structured data storage with version control
✅ Canonical key deduplication (YouTube, Bilibili, URLs, notes)
✅ Dual-track authentication (OAuth + PAT)
✅ Real-time draft auto-save
✅ Edge deployment with global low latency
✅ Responsive UI (mobile-friendly)

Kord serves as a lightweight, versioned storage backend - whether you’re archiving video subtitles, tuning AI prompts, or storing application data. Give it a try at kord.sedationh.top.