SegmentState you can address.

A type-safe state engine for Octane and React built around structural paths, targeted subscriptions, and atomic commits.

Get started

View on GitHub

Live storerunning Segment core

interactive

1const score = s.rows.at('beta').score

2store.observe(score, render)

3store.update(score, n => n + 1)

rows/alpha/score

rows/beta/score

rows/gamma/score

Choose an address to commitrows/beta/score

Structural paths

Reach declared state through typed refs or stable path strings such as users/42/profile/name.

Targeted updates

Wake observers of the affected address instead of evaluating every subscriber on every write.

Atomic commits

Publish several writes together, attribute them to one source, and roll them all back if the transaction throws.

O(observed) memory

Large keyed segments materialize nodes for watched addresses and prune them after the last observer leaves.

Async state built in

Resources share addressing with local state and add caching, cancellation, staleness, live values, and save-back.

Runtime independent

Use the DOM-free core from UI code, workers, tests, sockets, persistence layers, or server rendering.

One model from state to transport

import { createStore, segment } from 'segment-state';

const store = createStore({
	todos: segment({ title: '', completed: false }),
});

const completed = store.state.todos.at('docs').completed;

store.observe(completed, () => console.log(store.get(completed)));
store.set(completed, true, 'todo/complete');

A ref identifies state without exposing a mutable object. The same address can be observed by a component, written inside a transaction, serialized for hydration, or matched by an external port.

Measured behavior

Segment benchmark comparison against Valtio, Jotai, Redux, and Zustand

The benchmark suite prioritizes deterministic counts—callbacks woken, selector runs, and bookkeeping retained after churn—then measures write, memory, and mount costs in isolated processes. Timing and heap results are machine-specific.

Read the methodology and reproduce the run

Start with the shape of your data

The getting-started guide builds a store from plain values, shows how refs and transactions work, and introduces keyed segments. Continue with the state model or go directly to the advanced guide for resources, SSR, ports, and lifecycle guarantees. When you want to interact with the model first, open the browser playground.