@1gr14/error0

One typed, serializable Error class for errors that travel across your app.

Errors travel. You throw in one layer and catch in another. Sometimes it's your error, sometimes a native Error, sometimes an Axios or Zod error, sometimes just a string. error0 turns any of them into one typed class you control. You attach typed fields with small plugins, those fields flow up through cause chains, and the whole error serializes to JSON and back — so it survives a trip across a process, a queue, or the network.

import { Error0 } from '@1gr14/error0'
import { statusPlugin } from '@1gr14/error0/plugins/status'
import { codePlugin } from '@1gr14/error0/plugins/code'
import { metaPlugin } from '@1gr14/error0/plugins/meta'
import { causePlugin } from '@1gr14/error0/plugins/cause'
import { stackPlugin } from '@1gr14/error0/plugins/stack'
import { responsePlugin } from '@1gr14/error0/plugins/response'
import { redirectPlugin } from '@1gr14/error0/plugins/point0-redirect'
import { flatOriginalPlugin } from '@1gr14/error0/plugins/flat-original'
import { expectedPlugin } from '@1gr14/error0/plugins/expected'

// One error class for your whole app — compose built-in plugins and your own.
export const AppError = Error0.mark('AppError')
  .use(statusPlugin({ transport: 'public' }))
  .use(
    codePlugin({ codes: ['UNAUTHORIZED', 'FORBIDDEN'], transport: 'public' }),
  )
  .use(metaPlugin()) // private — lands in serializePrivate() only
  .use(causePlugin()) // the cause chain (Zod, Axios, ...) survives serializePrivate()
  .use(stackPlugin()) // the stack policy, explicit: private by default
  .use(responsePlugin())
  .use(redirectPlugin())
  .use(flatOriginalPlugin())
  .use(expectedPlugin())
  .use(betterAuthErrorPlugin) // ← your own plugin, composed like the rest

// build errors with typed fields
const inner = new AppError('Token expired', {
  status: 401,
  code: 'UNAUTHORIZED',
})

// stack them — wrap a cause, and fields flow up the chain
const outer = new AppError('Request failed', { cause: inner })
outer.status // 401  ← flowed up from the inner cause
outer.flow('status') // [undefined, 401]  — the value at each level of the chain

// coerce anything at a boundary, then serialize a client-safe payload
const json = AppError.serializePublic(outer) // public fields only; no stack, no meta

// ...and rebuild a real AppError on the other side
const restored = AppError.from(json)
restored.status // 401  ← survived the round-trip
restored.code // 'UNAUTHORIZED'

Install

bun add @1gr14/error0
# or: npm install / pnpm add / yarn add

Bun 1+ or Node.js 20+. ESM only.

Any error becomes your error

Start here, because this is the problem error0 was built for. You catch an unknown. You want a typed error you can trust. Error0.from() gives you one, every time:

import { Error0 } from '@1gr14/error0'

Error0.from(new Error('boom')) // wraps the native error, keeps it as `cause`
Error0.from('boom') // wraps the string
Error0.from({ message: 'boom' }) // rebuilds from a serialized object
Error0.from(error0Instance) // already an Error0 → returned as-is

try {
  await doStuff()
} catch (e) {
  throw Error0.from(e) // always an Error0, original preserved as `cause`
}

Error0 is a real subclass of Error, so everything you expect still works:

const err = new Error0('nope')
err instanceof Error0 // true
err instanceof Error // true
err.message // "nope"
err.stack // present

Prefer one error class

You usually want a single AppError for the whole app — not a DbError, ApiError, ValidationError zoo. Model the differences as fields, not classes. A field can hold anything — a whole object, not just a primitive — and you choose whether it crosses the wire.

// Don't reach for a separate DbError — add a field holding the raw driver error
const AppError = Error0.use('prop', 'dbError', {
  init: (error: PostgresError) => error, // the input can be a whole object
  resolve: ({ flow }) => flow.find(Boolean),
  serialize: false, // keep it server-side; never send it to a client
  deserialize: false,
})

const err = new AppError('Query failed', { dbError: pgError })
err.dbError // the full driver error, typed — for your logs
AppError.serialize(err) // { message } — `dbError` never crosses the wire

One class to catch, one is(), one serialize contract — every concern lives as a typed field on it. The next sections show how fields work.

Give your errors typed fields

A bare message isn't enough. You want an HTTP status, a code, whatever your app needs. Add a field with .use('prop', ...). A field is four small functions, and each one exists for a reason:

const AppError = Error0.use('prop', 'status', {
  // init: declares the input type (here: number); can also transform it
  init: (input: number) => input,
  // resolve: builds err.status from `flow` (this error's value + all causes')
  resolve: ({ flow }) => flow.find(Boolean),
  // serialize: turn the value into JSON
  serialize: ({ resolved }) => resolved,
  // deserialize: read the value back when rebuilding from JSON
  deserialize: ({ value }) => (typeof value === 'number' ? value : undefined),
})

const err = new AppError('User not found', { status: 404 })
err.status // 404  ← typed as number | undefined

• init mainly declares the input type. Writing (input: number) is what makes new AppError('...', { status }) expect a number. You can transform here too (a status name → a number), but typing the input is the point.
• resolve decides what err.status returns. flow is the array of values down the cause chain — this error's own value plus every cause's, nearest first. flow.find(Boolean) means "the first one anyone set". More on this in the next section.
• serialize / deserialize are the two ends of the JSON boundary. No field crosses the wire without them.

Every field is optional on input. Even typed number, status can always be left out — it's then undefined. That convention is the trick behind Error0.from(): since no field is ever required, any error can become an Error0.

message and stack are reserved — adding them as props throws. To change how they serialize, use their own hooks: .use('message', { serialize }) and .use('stack', { serialize }) (the bundled stackPlugin, messageMergePlugin, and stackMergePlugin are built on these).

Fields flow through cause chains

Here's why resolve takes a flow. When you wrap an error, the inner error's status shouldn't vanish. flow is this error's value plus every cause's value, nearest first — so flow.find(Boolean) means "the first status anyone set":

const inner = new AppError('DB unreachable', { status: 503 })
const outer = new AppError('Could not load user', { cause: inner })

outer.status // 503  ← flowed up from `inner`
outer.flow('status') // [undefined, 503]
Error0.causes(outer, true) // [outer, inner] — the Error0 links in the chain

You decide the rule. Omit resolve (or resolve: false) and err.status is just this error's own value, ignoring causes. Return 500 and every error reports 500. The flow is yours to shape.

Add methods

Fields are data. You'll also want behavior — a question you ask an error often. Add a method:

const AppError = Error0.use('prop', 'status', {
  init: (input: number) => input,
  resolve: ({ flow }) => flow.find(Boolean),
  serialize: ({ resolved }) => resolved,
  deserialize: ({ value }) => (typeof value === 'number' ? value : undefined),
}).use(
  'method',
  'isStatus',
  (error, expected: number) => error.status === expected,
)

const err = new AppError('Forbidden', { status: 403 })
err.isStatus(403) // true

// every method is also a static that runs `from()` on its first argument —
// so it works on anything: an AppError, a serialized object, or a native error
AppError.isStatus(err, 403) // true

Adapt errors at construction

An adapt hook runs on every new error — including the ones from() builds out of foreign errors. It gets the live error, so it can read the cause, return fields to set them, and mutate native parts like message directly. This is where you teach Error0 to understand the rest of the world.

Default an unknown wrap to 500:

const ServerError = AppError.use('adapt', (error) => {
  // a native Error came in with no status → treat it as a server fault
  if (error.cause instanceof Error && error.status === undefined) {
    return { status: 500 } // returned fields are assigned to the error
  }
})

ServerError.from(new Error('socket hang up')).status // 500

Turn a ZodError into a clean 422 — status from the return value, message from the error's first issue:

import { z } from 'zod'

const ApiError = AppError.use('adapt', (error) => {
  if (error.cause instanceof z.ZodError) {
    // mutate `message` to rewrite it; return fields to set them
    error.message = error.cause.issues[0]?.message ?? error.message
    return { status: 422 }
  }
})

const err = ApiError.from(zodError) // a ZodError you caught upstream
err.message // 'Invalid email address'  ← first Zod issue
err.status // 422

Two levers, both shown above: return an object to set typed fields, and mutate the error for its native parts (message, stack).

Package fields into reusable plugins

Defining status inline once is fine. Defining it in every service is not. Wrap it in a plugin with Error0.plugin() and reuse it everywhere:

export const statusPlugin = () =>
  Error0.plugin().prop('status', {
    init: (input: number) => input,
    resolve: ({ flow }) => flow.find(Boolean),
    serialize: ({ resolved }) => resolved,
    deserialize: ({ value }) => (typeof value === 'number' ? value : undefined),
  })

const AppError = Error0.use(statusPlugin())

Each .use(...) returns a new class with the previous fields plus the new ones, all typed. Stack as many as you like:

const AppError = Error0.use(statusPlugin()).use(codePlugin())
const ApiError = AppError.use(tagsPlugin()) // keeps status + code, adds tags

Batteries included

The common fields are already written. Import only what you use, from @1gr14/error0/plugins/*:

import { Error0 } from '@1gr14/error0'
import { statusPlugin } from '@1gr14/error0/plugins/status'
import { codePlugin } from '@1gr14/error0/plugins/code'
import { tagsPlugin } from '@1gr14/error0/plugins/tags'

const AppError = Error0.use(statusPlugin())
  .use(codePlugin({ codes: ['NOT_FOUND', 'BAD_REQUEST'] as const }))
  .use(tagsPlugin({ tags: ['retryable', 'user-error'] as const }))

const err = new AppError('User not found', {
  status: 404,
  code: 'NOT_FOUND', // typed: only the codes you listed
  tags: ['user-error'], // typed: only the tags you listed
})

err.hasTag('user-error') // true  ← method from tagsPlugin

| Plugin | Adds | What it does | | ---------------------------------------------------- | ----------------------------------- | ------------------------------------------------------------------- | | statusPlugin | status: number | HTTP-style status, with optional enum and strict mode. | | codePlugin | code: string | Machine-readable code, with an optional whitelist. | | codeStatusPlugin | code: string, status: number | Both in one: a { CODE: status } map auto-fills the status. | | tagsPlugin | tags: string[], hasTag() | Dedup'd tags merged across the cause chain. | | metaPlugin | meta: Record<string, unknown> | JSON-safe metadata, merged across causes (nearest wins). | | expectedPlugin | expected: boolean, isExpected() | Flag errors that aren't bugs, so you don't log them as such. | | causePlugin | cause chain on serialize | Carry the cause chain: Error0s rebuilt, foreign errors kept as name/message/stack. | | headersPlugin | headers: Record<string, string> | Merge HTTP headers from the cause chain (not serialized). | | responsePlugin | response: Response | Attach a Response object (not serialized). | | stackPlugin | stack policy on serialize | The default stack gate as a plugin — transport picks who sees it. | | messageMergePlugin | merged message on serialize | Join the message chain when serializing. | | stackMergePlugin | merged stack on serialize | Join the stack chain when serializing. | | flatOriginalPlugin | adapt hook | Unwrap a native Error cause — adopt its message and stack. | | redirectPlugin | redirect | Attach a navigation redirect (for point0). |

Send an error across the wire

This is the payoff. Serialize to plain JSON, ship it anywhere, rebuild a real typed error on the other side:

const err = new AppError('User not found', { status: 404, code: 'NOT_FOUND' })

const json = err.serializePrivate() // full object, safe to JSON.stringify
const back = AppError.from(json) // a real AppError again

back instanceof AppError // true
back.status // 404
back.code // 'NOT_FOUND'

Public vs private

Some fields are for your logs, not your users. The two audiences have named methods: serializePublic() is what an untrusted client may see; serializePrivate() is the full view for trusted consumers — logs, dev tooling. Both are thin sugar over serialize(isPublic):

const AppError = Error0.use(statusPlugin({ transport: 'public' })) // visible to clients
  .use(codePlugin()) // transport: 'private' by default

const err = new AppError('Nope', { status: 403, code: 'FORBIDDEN' })

err.serializePublic() // { message, status }   ← no code, no stack
err.serializePrivate() // { message, status, code, stack }

Send err.serializePublic() to the browser, log err.serializePrivate() on the server.

Every bundled plugin takes the same transport option: 'public' — the field is in both outputs; 'private' — only in serializePrivate(); 'none' — never serialized at all.

There's no magic here — it's the field's own serialize. The function gets a call-time isPublic flag and decides what to return. Return a value and the field lands in the JSON; return undefined and it's dropped entirely. Here's the exact gate every bundled plugin uses:

// inside statusPlugin({ transport })  ← 'public' | 'private' | 'none', the plugin option
serialize: ({ resolved, isPublic }) => {
  // never serialized, or private on a public call → hide it
  if (transport === 'none' || (transport === 'private' && isPublic))
    return undefined
  return resolved // otherwise, put the value in the JSON
}

So the transport option is just the default for that gate. Write your own serialize and you decide exactly what crosses the wire — mask a value, round it, or drop it.

Recognize your errors with is (and mark)

One AppError is usually enough — model the rest as fields (see Prefer one error class). But if you do split into several classes, is() tells them apart, and narrows the type inside the branch:

const ApiError = Error0.use(statusPlugin())
const DbError = Error0.use(codePlugin())

try {
  await handler()
} catch (e) {
  if (ApiError.is(e)) {
    e.status // typed — `e` is an ApiError here
  } else if (DbError.is(e)) {
    e.code // typed — `e` is a DbError here
  }
}

is() checks instanceof under the hood, so distinct classes stay distinct — no setup needed. (Still, prefer one error class per project; reach for more only when it genuinely helps.)

When instanceof isn't enough — mark

instanceof breaks when the same class ships in two bundles (a server build and a client build, say) — the two copies are different classes. mark brands a class with a stable id that is() checks instead of the prototype chain, so recognition survives that boundary:

const ApiError = Error0.mark('myapp/api').use(statusPlugin())

ApiError.is(err) // matched by brand, even where `instanceof` would fail

Use a string or a Symbol.for('...') as the mark — both are stable across bundles. Never a plain Symbol('...'): it's unique per bundle. Give several classes the same mark and is() treats them as one family.

Better stack traces in dev

Bundlers (Vite, tsx, esbuild) rewrite your code, so stack traces point at compiled output instead of your source. error0 calls an optional global hook on every error and each of its causes at construction, so a tool can remap the stack. It's a no-op when NODE_ENV === 'production'.

Wire it once — for example, with Vite's SSR fixer:

// dev setup only
globalThis.__ERROR0_FIX_STACKTRACE__ = (error) =>
  viteDevServer.ssrFixStacktrace(error)

Now every Error0, and each error in its cause chain, gets readable, source-mapped stack traces in development.

API reference

Static

| Call | Result | | ------------------------------------ | -------------------------------------------------------------------------------------- | | Error0.use(plugin) | Extend with a plugin builder. | | Error0.use('prop', key, options) | Add one typed field. | | Error0.use('method', key, fn) | Add an instance method. | | Error0.use('adapt', fn) | Run a function on every new error. | | Error0.plugin() | Start a plugin builder. | | Error0.from(unknown) | Coerce anything into an Error0 instance. | | Error0.is(unknown) | Type guard for this class. | | Error0.serialize(error, isPublic?) | Serialize to a plain object (public by default). | | Error0.serializePublic(error) | What an untrusted client may see. | | Error0.serializePrivate(error) | The full view — for logs and dev tooling. | | Error0.round(error, isPublic?) | from(serialize(error)). | | Error0.causes(error) | The cause chain as an array. | | Error0.flow(error, key) | A field's values down the cause chain. | | Error0.assign(error, props) | Set fields on an existing error. | | Error0.mark(string \| symbol) | Brand the class for cross-bundle is() checks; a string mark also becomes err.name. | | Error0.MAX_CAUSES_DEPTH | Cap on cause-chain walks (default 99). |

Instance

| Call | Result | | -------------------------- | ------------------------------------------- | | err.serialize(isPublic?) | Serialize this error (public by default). | | err.serializePublic() | What an untrusted client may see. | | err.serializePrivate() | The full view — for logs and dev tooling. | | err.round(isPublic?) | Round-trip this error. | | err.assign(props) | Set fields, return this. | | err.flow(key) | A field's values down the cause chain. | | err.causes() | The cause chain as an array. | | err.own | Raw fields set on this error (pre-resolve). |

Plugin builder

Error0.plugin() starts a reusable builder. Each call returns a new builder with the types added; pass the finished builder to Error0.use(builder).

| Call | Result | | --------------------- | ------------------------------------------------------------ | | Error0.plugin() | Start a plugin builder. | | .prop(key, options) | Add a typed field (same options as Error0.use('prop', …)). | | .method(key, fn) | Add an instance method. | | .adapt(fn) | Add a hook that runs on every new error. | | .cause(value) | Customize how the cause serializes and rebuilds. | | .stack(value) | Customize how the stack serializes. | | .message(value) | Customize how the message serializes. | | .use(kind, …) | Same as the methods above, addressed by kind. | | .use(builder) | Merge another builder into this one. |

// `.prop(...)` is shorthand for `.use('prop', ...)` — both exist on the builder
const statusPlugin = () =>
  Error0.plugin().prop('status', {
    /* init / resolve / serialize / deserialize */
  })

Community

Questions, bugs, or want to hang with other builders? Join the 1gr14 community — one hub for all our open-source projects, this one included. Get help, share what you built, or just say hi: 1gr14.dev/community

Contributing

Issues and PRs welcome. See CONTRIBUTING.md and the Code of Conduct. Commits follow Conventional Commits. Security reports: SECURITY.md.

License

MIT

Building open-source software for the glory of the Lord Jesus Christ ☦️
With love for developers of all backgrounds around the world ❤️
Sergei Dmitriev, 2026 😎