Skip to content
fermion-patterns logo

Fermion Patterns

fermion-patterns

Schema-driven Atom.family patterns for TMNL. Covers Fermion builder API, registry patterns, algebra interpreters, and React integration. (project)

SKILL.md

Full skill instructions

Fermion Patterns

Schema-driven Atom.family library for TMNL. Fermion provides type-safe, memoized atom families with Effect-based CRUD operations.

The name "Fermion" comes from fundamental particles that obey the Pauli exclusion principle - no two fermions can share the same quantum state. This maps to Atom.family memoization where each key gets exactly one atom.


Decision Tree

Need per-entity reactive state?
├── Single entity type with ID?
│   └── Fermion.fromSchema(Schema).withKey("id")
├── Composite key (userId + orderId)?
│   └── Fermion.fromSchema(Schema).withCompositeKey(["userId", "orderId"])
├── Need async fetch on access?
│   └── .withFetch((key) => Effect.tryPromise(...))
├── Need persistence?
│   └── .withPersist((value) => Effect.tryPromise(...))
└── Testing / local state only?
    └── .provideLayer(makeMemoryAlgebra())

CRITICAL: Registry Pattern

This is the most common source of bugs with effect-atom and Fermion.

The Problem

Atom.set() and Atom.get() return Effects that require AtomRegistry context:

// Atom.set() returns Effect<void, never, AtomRegistry>
// Atom.get() returns Effect<A, never, AtomRegistry>

// WRONG - This does NOTHING! The Effect is never executed
Atom.set(myAtom, newValue)  // ❌ Returns Effect, doesn't mutate

// WRONG - This returns an Effect, not the value
const value = Atom.get(myAtom)  // ❌ Returns Effect<A>, not A

The Solution: Registry Singleton

Create a registry singleton and use registry.set() / registry.get() for synchronous mutations:

import { Registry, RegistryContext } from "@effect-atom/​atom-react"

// 1. Create singleton registry at module level
export const myRegistry = Registry.make()

// 2. Create Provider wrapper
export function MyRegistryProvider({ children }: { children: React.ReactNode }) {
  return (
    <RegistryContext.Provider value={myRegistry}>
      {children}
    </​RegistryContext.Provider>
  )
}

// 3. Use registry.set() / registry.get() for sync mutations
function handleUpdate(newValue: string) {
  // ✓ Synchronous - mutates immediately
  myRegistry.set(myAtom, newValue)

  // ✓ Synchronous - returns value immediately
  const current = myRegistry.get(myAtom)
}

When to Use Which

ContextPatternWhy
Inside Effect.gen()Atom.set() / Atom.get()Effect context provides AtomRegistry
Inside runtimeAtom.fn()ctx.set() / ctx.get()Context parameter provides access
React callback / event handlerregistry.set() / registry.get()Synchronous, no Effect needed
Outside any contextregistry.set() / registry.get()Only option for sync access

Complete Example: IoT Sensor Pattern

// atoms/​index.ts
import { Atom, Registry, RegistryContext } from "@effect-atom/​atom-react"
import * as Fermion from "@/​lib/​fermion"
import { Schema, Effect } from "effect"

// ─── Registry Singleton ───────────────────────────────────────
export const iotRegistry = Registry.make()

export function IoTRegistryProvider({ children }: { children: React.ReactNode }) {
  return React.createElement(
    RegistryContext.Provider,
    { value: iotRegistry },
    children
  )
}

// ─── Sensor Schema ────────────────────────────────────────────
const SensorSchema = Schema.Struct({
  id: Schema.String.pipe(Schema.brand("SensorId")),
  name: Schema.NonEmptyString,
  value: Schema.Number,
  unit: Schema.String,
  lastUpdate: Schema.DateFromSelf,
})
type Sensor = typeof SensorSchema.Type

// ─── Fermion Family ───────────────────────────────────────────
export const sensorFamily = Fermion.fromSchema(SensorSchema)
  .withKey("id")
  .withFetch((id) =>
    Effect.tryPromise(() =>
      fetch(`/​api/​sensors/​${id}`).then(r => r.json())
    )
  )
  .build()

// ─── Log Atom (for debugging) ─────────────────────────────────
export const iotLogAtom = Atom.make<readonly string[]>([])

// ─── Sync Mutation Helper ─────────────────────────────────────
export function logMessage(msg: string) {
  const timestamp = new Date().toISOString().slice(11, 19)
  const current = iotRegistry.get(iotLogAtom)
  // Keep last 15 messages
  iotRegistry.set(iotLogAtom, [...current.slice(-14), `${timestamp}: ${msg}`])
}
// Component.tsx
import { useAtomValue } from "@effect-atom/​atom-react"
import { sensorFamily, iotLogAtom, logMessage, IoTRegistryProvider } from "./​atoms"
import * as Result from "@effect-atom/​atom/​Result"

function SensorDisplay({ sensorId }: { sensorId: string }) {
  const sensorResult = useAtomValue(sensorFamily(sensorId))

  useEffect(() => {
    // Trigger fetch
    sensorFamily.fetch(sensorId).pipe(Effect.runPromise)
    logMessage(`Fetching sensor ${sensorId}`)
  }, [sensorId])

  return Result.match(sensorResult, {
    onInitial: () => <Loading />,
    onSuccess: (sensor) => (
      <div>
        <span>{sensor.name}</​span>
        <span>{sensor.value} {sensor.unit}</​span>
      </​div>
    ),
    onFailure: (error) => <Error error={error} />,
  })
}

function SensorLog() {
  const logs = useAtomValue(iotLogAtom)
  return (
    <pre className="text-xs">
      {logs.join("\n")}
    </​pre>
  )
}

// App must wrap with provider
function App() {
  return (
    <IoTRegistryProvider>
      <SensorDisplay sensorId="temp-001" />
      <SensorLog />
    </​IoTRegistryProvider>
  )
}

Builder API Reference

Entry Point: fromSchema()

import * as Fermion from "@/​lib/​fermion"
import { Schema } from "effect"

const UserSchema = Schema.Struct({
  id: Schema.String.pipe(Schema.brand("UserId")),
  name: Schema.NonEmptyString,
  email: Schema.String,
})

const userFamily = Fermion.fromSchema(UserSchema)
  .withKey("id")
  .withFetch(...)
  .build()

Key Configuration

Single Key: .withKey(field)
// Key type inferred from schema field
const family = Fermion.fromSchema(UserSchema)
  .withKey("id")  // Key type: string & Brand<"UserId">
Composite Key: .withCompositeKey([...fields])
const OrderSchema = Schema.Struct({
  userId: Schema.String.pipe(Schema.brand("UserId")),
  orderId: Schema.String.pipe(Schema.brand("OrderId")),
  items: Schema.Array(OrderItemSchema),
})

const orderFamily = Fermion.fromSchema(OrderSchema)
  .withCompositeKey(["userId", "orderId"])
  // Key type: { readonly userId: string; readonly orderId: string }

// Usage
const orderAtom = orderFamily({ userId: "u-1", orderId: "o-42" })

Operations

.withFetch() - Required
.withFetch((id) =>
  Effect.tryPromise(() =>
    fetch(`/​api/​users/​${id}`).then(r => r.json())
  )
)

// With service dependency
.withFetch((id) =>
  Effect.gen(function* () {
    const api = yield* UserApi
    return yield* api.fetchUser(id)
  })
)
.withPersist() - Optional
.withPersist((user) =>
  Effect.tryPromise(() =>
    fetch(`/​api/​users/​${user.id}`, {
      method: "PUT",
      body: JSON.stringify(user)
    })
  )
)
.withRemove() - Optional
.withRemove((id) =>
  Effect.tryPromise(() =>
    fetch(`/​api/​users/​${id}`, { method: "DELETE" })
  )
)

Lifecycle Hooks

.withBeforeFetch()
.withBeforeFetch((id) =>
  Effect.log(`Fetching user ${id}`)
)
.withAfterFetch()
.withAfterFetch((id, user) =>
  Effect.gen(function* () {
    yield* Effect.log(`Fetched user ${user.name}`)
    // Can transform the value
    return { ...user, fetchedAt: new Date() }
  })
)

Dependency Injection

.provideLayer()
.provideLayer(HttpClient.layer)
.provideLayer(UserCacheLive)
.provideService()
.provideService(UserCache, {
  get: (id) => Effect.succeed(undefined),
  set: (id, user) => Effect.void,
})

Configuration

.withLifecycle()
// Entity cache - persist indefinitely (default)
.withLifecycle({ keepAlive: true })

// Transient data - reset when no subscribers
.withLifecycle({ keepAlive: false })

// Time-limited cache
.withLifecycle({ keepAlive: true, ttl: Duration.minutes(5) })
.withTTL() - Shorthand
// Equivalent to .withLifecycle({ keepAlive: true, ttl })
.withTTL(Duration.minutes(5))
.withReactivity() - Fine-grained subscriptions
.withReactivity((key, user) => [user.email, user.role])

Terminal Operations

.build() - All dependencies satisfied
// Type-safe: fails to compile if R is not `never`
const family = Fermion.fromSchema(UserSchema)
  .withKey("id")
  .withFetch(...)
  .provideLayer(HttpClient.layer)  // Satisfies HttpClient requirement
  .build()  // ✓ Compiles
.buildWithDeps() - Runtime dependency injection
// Use when layers will be provided at runtime via Atom.runtime
const family = Fermion.fromSchema(UserSchema)
  .withKey("id")
  .withFetch((id) => Effect.flatMap(UserApi, api => api.fetch(id)))
  .buildWithDeps()  // R = UserApi

// Provide at runtime
const runtimeAtom = Atom.runtime(UserApiLive)

Fermion Interface

Once built, a Fermion family provides:

interface Fermion<A, I, E, R, K> {
  // Callable - get atom for key (memoized)
  (key: K): Atom<Result<A, E>>

  // Same as callable
  atomFor(key: K): Atom<Result<A, E>>

  // Effectful operations (require AtomRegistry)
  fetch(key: K): Effect<A, E, R | AtomRegistry>
  persist(value: A): Effect<void, E, R | AtomRegistry>
  remove(key: K): Effect<void, E, R | AtomRegistry>

  // Utilities
  invalidate(key: K): Effect<void, never, AtomRegistry>
  prefetch(keys: readonly K[]): Effect<void, E, R | AtomRegistry>

  // Metadata
  schema: Schema<A, I, R>
  keyField: string | readonly string[]
}

Usage

// Get atom (memoized - same key always returns same atom)
const userAtom = userFamily("user-123")
const sameAtom = userFamily("user-123")  // Same reference

// React subscription
const userResult = useAtomValue(userFamily(userId))

// Trigger fetch
await userFamily.fetch(userId).pipe(Effect.runPromise)

// Persist changes
await userFamily.persist(updatedUser).pipe(Effect.runPromise)

// Remove entity
await userFamily.remove(userId).pipe(Effect.runPromise)

// Invalidate (force refetch on next access)
await userFamily.invalidate(userId).pipe(Effect.runPromise)

// Prefetch multiple
await userFamily.prefetch(["u-1", "u-2", "u-3"]).pipe(Effect.runPromise)

Algebra Patterns

What is FermionAlgebra?

The algebra defines the data operations (fetch, persist, remove) that power a Fermion family. Different interpreters implement the same interface for different backends.

interface FermionAlgebra<A, E, R, K> {
  readonly fetch: (key: K) => Effect<A, E, R>
  readonly persist?: (value: A) => Effect<void, E, R>
  readonly remove?: (key: K) => Effect<void, E, R>
  readonly beforeFetch?: (key: K) => Effect<void, E, R>
  readonly afterFetch?: (key: K, value: A) => Effect<A, E, R>
}

Built-in Interpreters

fromFunctions() - Effect-based CRUD
import { fromFunctions } from "@/​lib/​fermion"

const userAlgebra = fromFunctions({
  fetch: (id: string) =>
    Effect.tryPromise(() => api.getUser(id)),
  persist: (user: User) =>
    Effect.tryPromise(() => api.updateUser(user)),
  remove: (id: string) =>
    Effect.tryPromise(() => api.deleteUser(id)),
})
fromFetch() - HTTP endpoints
import { fromFetch } from "@/​lib/​fermion"

const userAlgebra = fromFetch({
  baseUrl: "/​api/​users",
  // GET /​api/​users/​{id}
  // PUT /​api/​users/​{id}
  // DELETE /​api/​users/​{id}
})
makeMemoryAlgebra() - Testing / local state
import { makeMemoryAlgebra } from "@/​lib/​fermion"

// Creates in-memory storage with full CRUD
const testAlgebra = makeMemoryAlgebra<User, string>()

// Usage in tests
const family = Fermion.fromSchema(UserSchema)
  .withKey("id")
  .withFetch(testAlgebra.fetch)
  .withPersist(testAlgebra.persist)
  .withRemove(testAlgebra.remove)
  .build()

Composing Algebras

composeAlgebra() - Merge operations
import { composeAlgebra } from "@/​lib/​fermion"

const cacheAlgebra = fromFunctions({ fetch: cacheGet })
const apiAlgebra = fromFunctions({ fetch: apiFetch })

// Cache-then-API pattern
const composedAlgebra = composeAlgebra(cacheAlgebra, apiAlgebra)
withHooks() - Add lifecycle hooks
import { withHooks } from "@/​lib/​fermion"

const hookedAlgebra = withHooks(baseAlgebra, {
  beforeFetch: (key) => Effect.log(`Fetching ${key}`),
  afterFetch: (key, value) => {
    analytics.track("entity_fetched", { key })
    return Effect.succeed(value)
  },
})

React Integration

Basic Pattern

import { useAtomValue } from "@effect-atom/​atom-react"
import * as Result from "@effect-atom/​atom/​Result"
import { userFamily } from "./​atoms"

function UserProfile({ userId }: { userId: string }) {
  const userResult = useAtomValue(userFamily(userId))

  // Trigger fetch on mount
  useEffect(() => {
    userFamily.fetch(userId).pipe(Effect.runPromise)
  }, [userId])

  return Result.match(userResult, {
    onInitial: () => <Skeleton />,
    onSuccess: (user) => <ProfileCard user={user} />,
    onFailure: (error) => <ErrorBanner error={error} />,
  })
}

With Loading States

function UserProfile({ userId }: { userId: string }) {
  const userResult = useAtomValue(userFamily(userId))
  const [isRefreshing, setIsRefreshing] = useState(false)

  const handleRefresh = async () => {
    setIsRefreshing(true)
    await userFamily.fetch(userId).pipe(Effect.runPromise)
    setIsRefreshing(false)
  }

  return (
    <div>
      {Result.isSuccess(userResult) && (
        <ProfileCard user={userResult.value} />
      )}
      <button onClick={handleRefresh} disabled={isRefreshing}>
        {isRefreshing ? "Refreshing..." : "Refresh"}
      </​button>
    </​div>
  )
}

Optimistic Updates

function EditableUser({ userId }: { userId: string }) {
  const userResult = useAtomValue(userFamily(userId))

  const handleUpdate = async (newName: string) => {
    if (!Result.isSuccess(userResult)) return

    const user = userResult.value
    const updated = { ...user, name: newName }

    // Optimistic update via registry
    myRegistry.set(userFamily(userId), Result.success(updated))

    try {
      await userFamily.persist(updated).pipe(Effect.runPromise)
    } catch {
      // Rollback on failure
      myRegistry.set(userFamily(userId), Result.success(user))
    }
  }

  // ...
}

Testing Patterns

Isolated Registry

import { Registry } from "@effect-atom/​atom-react"
import { makeMemoryAlgebra } from "@/​lib/​fermion"

describe("UserFamily", () => {
  let testRegistry: Registry
  let testFamily: Fermion<User, ...>

  beforeEach(() => {
    testRegistry = Registry.make()

    const memoryAlgebra = makeMemoryAlgebra<User, string>()
    testFamily = Fermion.fromSchema(UserSchema)
      .withKey("id")
      .withFetch(memoryAlgebra.fetch)
      .withPersist(memoryAlgebra.persist)
      .build()
  })

  it("fetches user by id", async () => {
    // Seed data
    await testFamily.persist({
      id: "u-1",
      name: "Alice",
      email: "[email protected]",
    }).pipe(Effect.runPromise)

    // Fetch
    const user = await testFamily.fetch("u-1").pipe(Effect.runPromise)

    expect(user.name).toBe("Alice")
  })

  it("updates atom on fetch", async () => {
    await testFamily.fetch("u-1").pipe(Effect.runPromise)

    const atomValue = testRegistry.get(testFamily("u-1"))
    expect(Result.isSuccess(atomValue)).toBe(true)
  })
})

Snapshot Testing

it("matches family state snapshot", () => {
  // ... setup and operations

  const state = testRegistry.get(testFamily("u-1"))
  expect(state).toMatchSnapshot()
})

Anti-Patterns

ANTIPATTERN:ATOM_SET_WITHOUT_REGISTRY

// ❌ WRONG - Returns Effect, never executed
function handleClick() {
  Atom.set(myAtom, newValue)  // Does nothing!
}

// ✓ CORRECT - Use registry for sync mutations
function handleClick() {
  myRegistry.set(myAtom, newValue)  // Mutates immediately
}

ANTIPATTERN:FAMILY_IN_COMPONENT

// ❌ WRONG - Creates new family on every render
function Bad({ schema }) {
  const family = Fermion.fromSchema(schema).withKey("id").build()
  return <Child family={family} />
}

// ✓ CORRECT - Define at module level
const userFamily = Fermion.fromSchema(UserSchema).withKey("id").build()

function Good() {
  return <Child family={userFamily} />
}

ANTIPATTERN:MISSING_PROVIDER

// ❌ WRONG - No registry context, useAtomValue may not work
function App() {
  return <UserProfile userId="123" />
}

// ✓ CORRECT - Wrap with registry provider
function App() {
  return (
    <MyRegistryProvider>
      <UserProfile userId="123" />
    </​MyRegistryProvider>
  )
}

ANTIPATTERN:FORGETTING_EFFECT_EXECUTION

// ❌ WRONG - Effect returned but never run
function handleFetch() {
  userFamily.fetch(userId)  // Returns Effect, doesn't execute!
}

// ✓ CORRECT - Run the Effect
function handleFetch() {
  userFamily.fetch(userId).pipe(Effect.runPromise)
}

// ✓ ALSO CORRECT - In Effect context
const program = Effect.gen(function* () {
  const user = yield* userFamily.fetch(userId)
})

Canonical Files

FilePurpose
src/​lib/​fermion/​index.tsPublic API exports
src/​lib/​fermion/​Fermion.tsBuilder implementation
src/​lib/​fermion/​types.tsType definitions
src/​lib/​fermion/​algebra/Algebra interface and composition
src/​lib/​fermion/​interpreters/Built-in interpreters (effect, memory)
src/​components/​testbed/​FermionTestbed.tsxUsage examples with registry pattern

Advanced Atom.family Patterns

Keying Strategies

String Keys (Simple)
const userFamily = Atom.family((id: string) => Atom.make<User | null>(null))
Composite Keys (Tuple)
// For multi-part keys, use object or tuple
const orderFamily = Atom.family(
  ({ userId, orderId }: { userId: string; orderId: string }) =>
    Atom.make<Order | null>(null)
)

// Usage
const orderAtom = orderFamily({ userId: "u-1", orderId: "o-42" })
Hash-based Keys (Complex Objects)
import { Hash } from "effect"

// When key is complex object, provide hash function
const queryFamily = Atom.family(
  (query: SearchQuery) => Atom.make<SearchResult | null>(null),
  {
    // Custom equality for memoization
    isEqual: (a, b) => Hash.hash(a) === Hash.hash(b)
  }
)

Garbage Collection

Atom.family uses WeakRef + FinalizationRegistry when available:

// Atoms are garbage collected when:
// 1. No subscribers (React components)
// 2. Not keepAlive
// 3. No strong references

// Force keepAlive for cache scenarios
const cachedFamily = Atom.family((id: string) =>
  Atom.make<Data | null>(null).pipe(Atom.keepAlive)
)

// Explicit cleanup
registry.dispose(userFamily("u-123"))

Cache Invalidation

// Invalidate single entry
registry.refresh(userFamily("u-123"))

// Invalidate with Reactivity
const userAtom = Atom.make(() => fetchUser()).pipe(
  Atom.withReactivity(["users", "user:123"])
)

// Invalidate by key
runtimeAtom.fn(Effect.gen(function* () {
  yield* updateUser(data)
  yield* Reactivity.invalidate("user:123")
}))

Atom.runtime Patterns

Basic Runtime Setup

import { Atom } from "@effect-atom/​atom"
import { Layer } from "effect"

// Create runtime with service layer
const appRuntime = Atom.runtime(
  Layer.mergeAll(
    DatabaseService.Default,
    CacheService.Default,
    LoggingService.Default
  )
)

// Effectful atom using services
const usersAtom = appRuntime.atom(
  Effect.gen(function* () {
    const db = yield* DatabaseService
    return yield* db.query("SELECT * FROM users")
  })
)

Global Layers

// Add cross-cutting concerns globally
Atom.runtime.addGlobalLayer(
  Layer.setTracerEnabled(true)
)

Atom.runtime.addGlobalLayer(
  Layer.setConfigProvider(ConfigProvider.fromEnv())
)

Runtime Functions (runtimeAtom.fn)

// Create effectful operations callable from React
const ops = {
  fetchUser: appRuntime.fn<string>()(
    (userId) => Effect.gen(function* () {
      const api = yield* UserApi
      return yield* api.fetch(userId)
    })
  ),

  updateUser: appRuntime.fn<User>()(
    (user, ctx) => Effect.gen(function* () {
      yield* UserApi.update(user)
      // ctx.set for sync atom updates within Effect
      ctx.set(userAtom, Result.success(user))
    })
  )
}

// Usage in React
const handleSave = () => {
  registry.set(ops.updateUser, updatedUser)
}

Stream-backed Atoms (Atom.pull)

// Paginated/​infinite data
const messagesAtom = Atom.pull(
  Stream.paginateChunkEffect(0, (page) =>
    Effect.gen(function* () {
      const api = yield* MessagesApi
      const chunk = yield* api.getPage(page)
      const next = chunk.hasMore ? Option.some(page + 1) : Option.none()
      return [chunk.messages, next]
    })
  )
)

// Usage: pull next chunk
registry.set(messagesAtom, void 0) // Triggers next chunk

SubscriptionRef Integration

// Backed by SubscriptionRef for external updates
const priceAtom = appRuntime.subscriptionRef(
  Effect.gen(function* () {
    const ws = yield* WebSocketService
    return yield* ws.subscribe("prices")
  })
)

Result Handling Patterns

The Result Type

import * as Result from "@effect-atom/​atom/​Result"

type Result<A, E = never> =
  | { _tag: "Initial" }                    // Never fetched
  | { _tag: "Waiting"; previous?: A }      // Fetching (with optional stale data)
  | { _tag: "Success"; value: A }          // Success
  | { _tag: "Failure"; cause: Cause<E> }   // Failed

Pattern Matching

// Full match
Result.match(result, {
  onInitial: () => <Skeleton />,
  onWaiting: (previous) => previous
    ? <StaleData data={previous} isRefreshing />
    : <Loading />,
  onSuccess: (data) => <Content data={data} />,
  onFailure: (cause) => <Error cause={cause} />
})

// Guards
if (Result.isSuccess(result)) {
  console.log(result.value)
}
if (Result.isFailure(result)) {
  console.log(Cause.pretty(result.cause))
}

Derived Results

// Map over success
const nameResult = Result.map(userResult, user => user.name)

// FlatMap for dependent fetches
const profileResult = Result.flatMap(userResult, user =>
  registry.get(profileFamily(user.profileId))
)

Stale-While-Revalidate

function useStaleWhileRevalidate<A>(atom: Atom<Result<A>>) {
  const result = useAtomValue(atom)

  // Show previous data while revalidating
  const data = Result.match(result, {
    onInitial: () => null,
    onWaiting: (prev) => prev ?? null,
    onSuccess: (v) => v,
    onFailure: () => null
  })

  const isLoading = Result.isWaiting(result)
  const error = Result.isFailure(result) ? result.cause : null

  return { data, isLoading, error }
}

Domain Pattern Generalizations

Pattern: IoT Sensor Data

// Keyed by sensor ID, with polling refresh
const SensorSchema = Schema.Struct({
  id: Schema.String.pipe(Schema.brand("SensorId")),
  value: Schema.Number,
  unit: Schema.String,
  timestamp: Schema.DateFromSelf
})

const sensorFamily = Fermion.fromSchema(SensorSchema)
  .withKey("id")
  .withFetch((id) => SensorApi.read(id))
  .withTTL(Duration.seconds(5))  // Auto-refresh every 5s
  .build()

// Aggregate atom for dashboard
const sensorSummaryAtom = Atom.make((get) => {
  const sensors = ["temp-1", "humidity-1", "pressure-1"]
  const results = sensors.map(id => get.result(sensorFamily(id)))

  if (results.some(Result.isFailure)) {
    return { status: "error" }
  }
  if (results.some(r => !Result.isSuccess(r))) {
    return { status: "loading" }
  }

  const values = results.map(r => (r as Result.Success<Sensor>).value)
  return {
    status: "ready",
    avgTemp: values.find(s => s.id.startsWith("temp"))?.value,
    humidity: values.find(s => s.id.startsWith("humidity"))?.value
  }
})

Pattern: Financial Tick Data

// High-frequency updates with batching
const tickAtom = Atom.make<Tick | null>(null).pipe(Atom.keepAlive)
const tickHistoryAtom = Atom.make<Tick[]>([]).pipe(Atom.keepAlive)

// Batch rapid updates
const processTick = (tick: Tick) => {
  Atom.batch(() => {
    registry.set(tickAtom, tick)
    const history = registry.get(tickHistoryAtom)
    registry.set(tickHistoryAtom, [...history.slice(-99), tick])
  })
}

// Stream-backed for WebSocket
const liveTicksAtom = appRuntime.atom(
  Effect.gen(function* () {
    const ws = yield* TradingWebSocket
    return yield* ws.subscribe("ticks")
  })
)

Pattern: Chat/​Messaging

// Conversation family keyed by chat ID
const ConversationSchema = Schema.Struct({
  chatId: Schema.String.pipe(Schema.brand("ChatId")),
  messages: Schema.Array(MessageSchema),
  participants: Schema.Array(Schema.String)
})

const conversationFamily = Fermion.fromSchema(ConversationSchema)
  .withKey("chatId")
  .withFetch((id) => ChatApi.getConversation(id))
  .withPersist((conv) => ChatApi.updateConversation(conv))
  .build()

// Optimistic message append
const sendMessage = (chatId: string, content: string) => {
  const convAtom = conversationFamily(chatId)
  const result = registry.get(convAtom)

  if (Result.isSuccess(result)) {
    const conv = result.value
    const optimisticMsg = { id: crypto.randomUUID(), content, pending: true }

    // Optimistic update
    registry.set(convAtom, Result.success({
      ...conv,
      messages: [...conv.messages, optimisticMsg]
    }))

    // Actual send
    ChatApi.sendMessage(chatId, content).pipe(
      Effect.tap(() => conversationFamily.fetch(chatId)),
      Effect.runPromise
    )
  }
}

Pattern: Form State

// Field atoms for granular reactivity
const formFieldAtom = Atom.family((field: string) =>
  Atom.make({ value: "", error: null as string | null, touched: false })
)

// Validation derived atom
const formValidAtom = Atom.make((get) => {
  const fields = ["email", "password", "name"]
  return fields.every(f => {
    const field = get(formFieldAtom(f))
    return field.touched && !field.error
  })
})

// Batch field updates
const updateField = (name: string, value: string) => {
  Atom.batch(() => {
    const field = registry.get(formFieldAtom(name))
    const error = validateField(name, value)
    registry.set(formFieldAtom(name), {
      value,
      error,
      touched: true
    })
  })
}

// Submit with all fields
const submitForm = () => {
  const fields = ["email", "password", "name"]
  const data = Object.fromEntries(
    fields.map(f => [f, registry.get(formFieldAtom(f)).value])
  )
  return FormApi.submit(data)
}

Pattern: Data Grid State

// Grid state atoms
const gridSelectionAtom = Atom.make<Set<string>>(new Set())
const gridSortAtom = Atom.make<{ column: string; dir: "asc" | "desc" } | null>(null)
const gridPageAtom = Atom.make({ page: 0, pageSize: 50 })

// Row data family
const gridRowFamily = Atom.family((id: string) =>
  Atom.make<GridRow | null>(null)
)

// Derived: visible rows
const visibleRowsAtom = Atom.make((get) => {
  const { page, pageSize } = get(gridPageAtom)
  const sort = get(gridSortAtom)
  const allIds = get(allRowIdsAtom)

  let ids = [...allIds]
  if (sort) {
    ids.sort((a, b) => {
      const rowA = get(gridRowFamily(a))
      const rowB = get(gridRowFamily(b))
      // Sort logic...
    })
  }

  return ids.slice(page * pageSize, (page + 1) * pageSize)
})

// Batch selection
const selectRows = (ids: string[]) => {
  Atom.batch(() => {
    const current = registry.get(gridSelectionAtom)
    registry.set(gridSelectionAtom, new Set([...current, ...ids]))
  })
}

Keyphrase Triggers

This skill responds to:

  • [EFFECT:ATOM:SYNC] - registry.set/​get patterns
  • [EFFECT:ATOM:EFFECT] - Atom.set/​get inside Effect.gen
  • [EFFECT:ATOM:FAMILY] - Atom.family patterns
  • [EFFECT:ATOM:RUNTIME] - Atom.runtime patterns
  • [EFFECT:ATOM:RESULT] - Result type handling
  • [EFFECT:ATOM:BATCH] - Atom.batch patterns

Related Skills

  • /​effect-atom-integration - Atom.runtime, Atom.make, operation atoms
  • /​effect-schema-mastery - Schema.TaggedStruct, branded types
  • /​effect-service-authoring - Effect.Service patterns for algebras
  • /​effect-patterns - General Effect-TS patterns

More skills from majiayu000

xiaohongshu logo
majiayu000/claude-arsenal

xiaohongshu

xiaohongshu

286 148
View
agent-task-conductor logo
majiayu000/claude-skill-registry

agent-task-conductor

Conduct multi-agent task orchestration and workflow coordination.

663 1
View
conductor-setup logo
majiayu000/claude-skill-registry

conductor-setup

Initialize project with Conductor artifacts (product definition,

663 1
View
animation-designer logo
majiayu000/claude-skill-registry

animation-designer

Expert in web animations, transitions, and motion design using Framer Motion and CSS

663 1
View
diagramming logo
majiayu000/claude-skill-registry

diagramming

Creates Mermaid and ASCII diagrams for flowcharts, architecture, ERDs, state machines, mindmaps, and more. Use when user mentions diagram, flowchart, mermaid, ASCII diagram, text diagram, terminal diagram, visualize, C4, mindmap, architecture diagram, sequence diagram, ERD, or needs visual docume...

663 1
View
h3-pg logo
majiayu000/claude-skill-registry-data

h3-pg

PostgreSQL bindings for H3 hexagonal grid system. Use when working with H3 cells in Postgres, including spatial indexing, geometry/geography integration, and raster analysis.

23 1
View
conductor-development logo
majiayu000/claude-skill-registry

conductor-development

Context-Driven Development skill for projects using Conductor. Use this skill when you detect a `conductor/` directory in the project, when working on tasks defined in a `plan.md` file, or when the user asks about tracks, specs, or plans. Automatically applies TDD workflow, tracks task completion...

663 1
View
conductor-status logo
majiayu000/claude-skill-registry

conductor-status

Display project status, active tracks, and next actions

663 1
View
dockerization logo
majiayu000/claude-skill-registry

dockerization

Official Stakpak application containerization standard operating procedure, a step-by-step guidline to properly dockerize applications. This is a rule book curated by the Stakpak Team.

663 1
View

Popular AI tools

Kaiber logo
Video

Kaiber

Generate, edit, and beat-sync AI video with leading models in one workspace.

Paid
View
Vimcal logo
Productivity

Vimcal

The world's fastest calendar for remote work

Free
View

Transform Your Design with AI Designer by ImgCreator.ai

Freemium
View
Akool AI logo
Content & writing

Akool AI

Revolutionizing Video Production with AI-Powered Creativity

Paid
View

Extend an image past the frame and let AI fill the new aspect ratio.

Freemium
View
StarByFace logo
Security

StarByFace

Discover your celebrity doppelgänger with StarByFace!

Free
View
C

ChainClarity explains 700+ crypto whitepapers in plain English, with layered summaries, comparisons, research tools, alerts, and a $4.99 Pro plan.

Freemium
View
Opus Clip logo
Coding & apps

Opus Clip

Opus.ai: Revolutionize Your Web Experience

Free
View