Kilat vs. other frameworks
Kilat is not a meta-framework. It’s an edge-native boilerplate — a skeleton that already works, wired with the guardrails a deployed app needs. Here’s how it compares to the alternatives, and the reasoning behind each choice.
Kilat vs Next.js / Nuxt / SvelteKit
Section titled “Kilat vs Next.js / Nuxt / SvelteKit”Meta-frameworks give you a batteries-included runtime with conventions, routing, and rendering abstractions. Kilat gives you the same outcome — a full-stack app with SSR, auth, and migrations — but as explicit, readable code you own, not a framework you depend on.
| Kilat | Meta-frameworks | |
|---|---|---|
| What it is | Edge-native boilerplate | Framework with conventions |
| Routing | Hono routes, inline handlers | File-based routing abstraction |
| Rendering | Inertia v3 + in-process SSR | Framework rendering pipeline |
| Database | D1, raw prepared statements | ORM / adapter layer |
| Deploy | wrangler deploy |
Platform-specific (Vercel, etc.) |
| Free to start | ✅ Yes — 100K req/day, 500 MB D1 | ⚠️ Vercel/Netlify free tier is 100K req/month |
The tradeoff: meta-frameworks handle more for you, but you learn their abstractions. Kilat handles less, but what’s there is plain TypeScript you can read end to end. When something breaks, you grep your own code — not framework internals.
Kilat vs Dulak
Section titled “Kilat vs Dulak”Dulak is the Bun version of the same philosophy. Same deliberately-boring, zero-ops ethos; different runtime forces different tradeoffs:
| Kilat | Dulak | |
|---|---|---|
| Runtime | Cloudflare Workers (serverless edge) | Bun (single process) |
| Database | D1 (async SQLite at the edge) | bun:sqlite (sync, local) |
| Password hashing | PBKDF2 / Web Crypto (100K cap) | argon2 |
| Build | esbuild (client + SSR) | Bun.build |
| Deploy | wrangler deploy (edge, 300+ locations) |
Docker / VPS |
| Config | wrangler.toml [vars] |
.env |
| Rate limiting | no-op stub (stateless isolates) | in-memory Map |
| Uploads | not included | tus-based uploads |
| Compression | Wrangler auto-compresses | custom compress middleware |
| CPU time limit | 10 ms (free) / 5 min (paid) | Unlimited |
| Request limit | 100K/day (free) | Unlimited |
| Database size | 500 MB (free) / 10 GB (paid) | Unlimited (disk-bound) |
Pick Kilat if you want zero ops and global edge latency. Pick Dulak if you want a single long-lived process, sync DB access, argon2, and no platform limits — unlimited requests, unlimited CPU time, unlimited database size.
Already running Kilat and hitting the free tier limits? The Migrating from Kilat to Dulak guide walks through every step — and includes a copy-paste prompt you can give your AI agent to automate the whole migration.
Kilat vs Laravel
Section titled “Kilat vs Laravel”Laravel is the gold standard for batteries-included PHP web apps. It’s also
a time capsule from 2011: install PHP, install Composer, install a web
server, install a database server, configure PHP-FPM, tune your worker
processes, set up Nginx, manage SSL certs, and then run php artisan serve. Kilat brings the same “it just works” ergonomics — auth, sessions,
CSRF, migrations, mail — to TypeScript on the edge, minus the archaeology:
| Kilat | Laravel | |
|---|---|---|
| Language | TypeScript | PHP |
| Runtime | Cloudflare Workers (edge) | PHP-FPM / Octane |
| Database | D1 (SQLite at the edge) | MySQL / Postgres |
| ORM | none — raw prepared statements | Eloquent |
| Deploy | wrangler deploy |
Server + web server + PHP + Composer |
| Latency | 300+ edge locations | Single region |
| Setup | npm create kilat → running |
Install PHP, Composer, web server, DB |
| Cold start | Sub-millisecond | ~250 ms (PHP-FPM worker spin-up) |
| Monthly cost (paid) | $5 (Workers Paid) | VPS ($5–20) + DB ($0–30) |
| Free to start | ✅ Yes — 100K req/day, 500 MB D1 | ❌ No — need a server + database from day one |
| Things to install | Bun (1 thing) | PHP + Composer + web server + DB server |
Laravel gives you more batteries — queues, jobs, broadcasting, Eloquent relationships, Blade components, a massive ecosystem. Kilat gives you fewer, but what’s there runs at the edge with zero ops. You can learn Laravel’s conventions, or you can read Kilat end to end in an afternoon.
Why no ORM
Section titled “Why no ORM”Every dependency is a liability. D1 is async SQLite with a small, stable API:
prepare().bind().first/all/run. Raw parameterized queries are obviously
correct, type-checkable, and add zero upgrade surface. When your queries fit
in one db.ts file — and they do for a long time — an ORM is overhead you
pay for every upgrade. Add one when you genuinely need it; Kilat doesn’t
prescribe it.
Why PBKDF2, not argon2
Section titled “Why PBKDF2, not argon2”Workers limits PBKDF2 to 100K iterations — the OWASP-recommended 600K
throws NotSupportedError. 100K is still OWASP-acceptable for
PBKDF2-HMAC-SHA256. argon2 needs node:crypto, which isn’t available on
Workers (and nodejs_compat doesn’t cover it). Web Crypto’s crypto.subtle
is the native option, so that’s what Kilat uses. Dulak, running on Bun, uses
argon2 — the runtime decides.
Why D1, not Postgres
Section titled “Why D1, not Postgres”D1 is SQLite at the edge: read replicas automatically serve from the nearest
location, writes go to the primary. For a starter, that’s the right default —
zero connection pooling, zero external service, zero cost on the free tier.
When you outgrow SQLite (heavy write concurrency, advanced types), bring your
own Postgres via Hyperdrive. Kilat doesn’t prescribe it; the db.ts boundary
makes swapping the data layer a contained change.
Next steps
Section titled “Next steps”- Philosophy — the principles behind these tradeoffs.
- Architecture overview — how the choices shape the code.
- Configuration — the full env var reference.