npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@elumixor/synced

v0.1.3

Published

Optimistic writes with one queue: local-first records, retries, reconciliation and undo, over any backend

Readme

@elumixor/synced

Optimistic writes, one queue, one rule:

what you see = what the server said + what it hasn't acknowledged yet

Every app that writes ends up building this: draw the change immediately, keep a copy on the device, send it, reconcile whatever comes back. Written per screen it is four subtly different implementations and a standing supply of race conditions. This is that, written once.

bun add @elumixor/synced      # or npm / pnpm

Framework-agnostic: reactivity and storage are adapters, so it runs against Svelte runes, a React store, or nothing at all in tests. The server half is a four-method interface, so records can live in Postgres, in Gmail, in a calendar — anywhere that can answer "what changed since".

The rule

The view is recomputed from the confirmed records with the pending writes replayed over the top, every time either half moves. Two things follow, and they're the reason this exists:

  • A response can't clobber an edit. The edit is re-applied after the response lands, so a pull arriving mid-keystroke doesn't need a dirty flag to guard it.
  • A rejection needs no rollback. The write stops being replayed, and the row is server truth again on the next frame.

Writes carry an id the client makes up, so a retry after a dropped connection is free — the server recognises the id and doesn't send the email, or make the calendar event, twice. That id is spent the moment the write leaves the device: a later edit to the same record queues behind it rather than being folded into it, because a mutation the server may already have logged would be answered "already applied" and the folded-in edit would never be read.

Signing out ends the round trip that was in flight with it. An answer to a request the previous account made is dropped rather than absorbed, and the device's copy goes with it — including keys written by a model the app no longer configures.

Client

const client = syncedClient({
  models: { categories: model<Category>()({}) },
  transport: (request) => api.sync.$post(request),
  storage: localStorageAdapter,
  signals: runeSignals,
});
await client.hydrate();

const categories = client.models.categories;
categories.all;                                   // reactive, ordered
categories.create({ name: "Deep work", ... });    // on screen before the request exists
categories.update(id, { name: "Focus" });         // patches, not whole records
categories.delete(id);

const one = categories.get(id);
if (one) one.name = "Focus";                      // same thing, written the way it reads

Reactivity and storage are adapters (Signals, SyncedStorage) so the core is plain TypeScript: $state or a store on the app side, localStorage on the web (synchronously, so last session's records are on screen in the first frame), Capacitor Preferences on a device, nothing at all in the tests.

Undo

Undo isn't a rollback — the server has usually applied the write already, and on a mailbox or a calendar it certainly has. So it's a new write in the opposite direction, worked out when the original was made:

client.transaction(() => {          // one press of undo, however many writes
  categories.update(id, { name });
  categories.update(id, { color });
});
client.undo();
client.redo();

Actions

Not every write is a field changing. An action is a write whose meaning belongs to the server — answering an invitation, archiving a Gmail thread, moving a Notion page — with the client saying only what the screen should show meanwhile, and what undoing it means:

model<Event>()({
  actions: {
    rsvp: {
      optimistic: (event, payload: { response: Response }) => ({ response: payload.response }),
      invert: (event, _payload) => ({ kind: "action", model: "events", id: event.id,
                                      name: "rsvp", payload: { response: event.response } }),
    },
  },
});

Server

A model says what its records look like, where they live, and what can be done to one. "Where they live" is an interface, not a table: pull, read, create, update, remove. Back it with your ORM, with somebody else's API, with anything that can answer "what changed since".

const categories = syncedModel({
  name: "categories",
  schema: categorySchema,
  store: drizzleCategories,
  writable: (category) => !category.isBuffer,     // server-managed rows refuse client writes
  actions: { archive: action(z.object({}), archiveCategory) },
});

const sync = new SyncServer([categories], mutationLog);

SyncServer.handle applies the batch in order, remembers the mutation ids it has carried out, and answers with everything that changed. A write it will never accept comes back named and reasoned, and the rest of the batch still lands. A write that failed for a reason a retry could fix isn't rejected — it throws, and the client tries the whole request again with its queue intact.

Tests

bun test — 100% of lines and functions, enforced (bunfig.toml), so an untested branch fails the run rather than showing up in a coverage report nobody reads. bun run test:watch while working; bun run test:coverage writes lcov to coverage/.

Two files: test/sync.test.ts is the behaviour the library promises — optimism, rejection, a response landing mid-edit, lost replies, offline bursts, restarts, tombstones, undo and redo. test/edges.test.ts is the half-happening cases, where being nearly right is worse than failing: a reply lost after the server applied the write, a sign-out under a request in flight, an answer describing a row older than the one already held, a tombstone arriving late, a client creating a row the server manages itself, a store that is down versus a store that says no.