- TypeScript 46.6%
- Clojure 44%
- JavaScript 2.7%
- Python 2.6%
- PLpgSQL 1.3%
- Other 2.7%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
|
Some checks failed
JS CI / detect_changes (push) Has been cancelled
JS CI / Build & Test packages (push) Has been cancelled
JS CI / Check formatting (push) Has been cancelled
Clojure CI / clj-check (0) (push) Has been skipped
Clojure CI / detect_changes (push) Successful in 17s
Clojure CI / clj-build-check (push) Has been skipped
Clojure CI / lint (push) Has been skipped
self-hosting/ci was one copy of a Forgejo Actions script I had been
copying from repo to repo, and the copies picked up fixes out of sync.
The script now lives in its own repo and is installed once per machine
as forgejo-ci-helper. It reads the same secrets.nuon. Run
forgejo-ci-helper wherever you ran self-hosting/ci.
bin/bootstrap is new. It checks that forgejo-ci-helper is on PATH and,
if it is not, says where to get it:
https://peoplesgrocers.com/code/GrocerPublishAgent/13.40-forgejo-ci-helper
trigger is a dry run unless --execute. A bare trigger dispatches the
workflows listed under default_workflows in secrets.nuon; otherwise
pass --workflow.
|
||
| .claude/commands | ||
| .github | ||
| bin | ||
| client | ||
| dev-llm-docs/prompts | ||
| examples | ||
| self-hosting | ||
| server | ||
| .envrc | ||
| .git-blame-ignore-revs | ||
| .gitignore | ||
| default.do | ||
| dev-llm-rules.md | ||
| LICENSE.md | ||
| Makefile | ||
| README.md | ||
| todo.plan | ||
| vercel.json | ||
Get Started · Examples · Try the Demo · Docs · Discord
Instant is the best backend for AI-coded apps. You get auth, permissions, storage, presence, and streams — everything you need to ship apps your users will love.
You write relational queries in the shape of the data you want and Instant handles all the data fetching, permission checking, and offline caching. When you change data, optimistic updates and rollbacks are handled for you as well. Plus, every query is multiplayer by default.
We also support ephemeral updates, like cursors, or who's online. Currently we have SDKs for Javascript, React, and React Native.
How does it look? Here's a barebones chat app in about 12 lines:
// ༼ つ ◕_◕ ༽つ Real-time Chat
// ----------------------------------
// * Updates instantly
// * Multiplayer
// * Works offline
import { init, id } from "@instantdb/react";
const db = init({
appId: process.env.NEXT_PUBLIC_APP_ID,
});
function Chat() {
// 1. Read
const { isLoading, error, data } = db.useQuery({
messages: {},
});
// 2. Write
const addMessage = (message) => {
db.transact(db.tx.messages[id()].update(message));
};
// 3. Render!
return <UI data={data} onAdd={addMessage} />;
}
Want to see for yourself? try a demo in your browser.
Motivation
Writing modern apps is full of schleps. Most of the time you start with the server: stand up databases, caches, ORMs, and endpoints. Then you write client-side code: stores, selectors, mutators. Finally you paint a screen. If you add multiplayer you need to think about stateful servers, and if you support offline mode, you need to think about IndexedDB and transaction queues.
To make things worse, whenever you add a new feature, you go through the same song and dance over and over again: add models, write endpoints, stores, selectors, and finally the UI.
Could it be better?
In 2021, we realized that most of the schleps we face as UI engineers are actually database problems in disguise. (We got into greater detail in this essay)
If you had a database on the client, you wouldn't need to think about stores, selectors, endpoints, or local caches: just write queries. If these queries were multiplayer by default, you wouldn't have to worry about stateful servers. And if your database supported rollback, you'd get optimistic updates for free.
So we built Instant. Instant gives you a database you can use in the client, so you can focus on what’s important: building a great UX for your users, and doing it quickly.
Architectural Overview
Here's how Instant works at a high level:
Under the hood, we store all user data as triples in one big Postgres database. A multi-tenant setup lets us offer a free tier that never pauses.
A sync server written in Clojure talks to Postgres. We wrote a query engine that understands datalog and InstaQL, a relational language that looks a lot like GraphQL:
// give me all users, their posts and comments
{
users: {
posts: {
comments: {
}
}
}
}
Taking inspiration from Asana’s WorldStore and Figma’s LiveGraph, we tail postgres’ WAL to detect novelty and invalidate relevant queries.
For the frontend, we wrote a client-side triple store. The SDK handles persisting a cache of recent queries to IndexedDB on web, and AsyncStorage in React Native.
All data goes through a permission system powered by Google's CEL library.
Getting Started
The easiest way to get started with Instant is by signing up on instantdb.com. You can create a functional app in 5 minutes or less.
If you have any questions, you can jump in on our discord.
Contributing
You can start by joining our discord and introducing yourself. Even if you don't contribute code, we always love feedback.
If you want to make changes, start by reading the client and server READMEs. There you'll find instructions to start Instant locally.