Tanstack Db

Overview

TanStack DB is a client-side embedded database layer built on differential dataflow. It maintains normalized collections, uses incremental computation for live queries, provides automatic optimistic mutations, and integrates with TanStack Query for data fetching. Sub-millisecond updates even with 100k+ rows.

Package: @tanstack/react-db Query Integration: @tanstack/query-db-collection Status: Beta (v0.5)

Installation

npm install @tanstack/react-db @tanstack/query-db-collection

Core Concepts

• Collections: Normalized data stores wrapping data sources (TanStack Query, Electric, etc.)
• Live Queries: Reactive subscriptions with SQL-like query builder
• Optimistic Mutations: Automatic instant UI updates with rollback on failure
• Differential Dataflow: Only recomputes affected query results on changes

Collections

Creating a Collection

import { createCollection } from '@tanstack/react-db'
import { queryCollectionOptions } from '@tanstack/query-db-collection'

const todoCollection = createCollection(
  queryCollectionOptions({
    queryKey: ['todos'],
    queryFn: async () => api.todos.getAll(),
    getKey: (item) => item.id,
    schema: todoSchema,
    onInsert: async ({ transaction }) => {
      await Promise.all(
        transaction.mutations.map((mutation) =>
          api.todos.create(mutation.modified)
        )
      )
    },
    onUpdate: async ({ transaction }) => {
      await Promise.all(
        transaction.mutations.map((mutation) =>
          api.todos.update(mutation.modified)
        )
      )
    },
    onDelete: async ({ transaction }) => {
      await Promise.all(
        transaction.mutations.map((mutation) =>
          api.todos.delete(mutation.original.id)
        )
      )
    },
  })
)

Sync Modes

// Eager (default): Load entire collection upfront. Best for 50k rows, search.
const largeCollection = createCollection(
  queryCollectionOptions({
    syncMode: 'on-demand',
    queryFn: async (ctx) => {
      const params = parseLoadSubsetOptions(ctx.meta?.loadSubsetOptions)
      return api.getProducts(params)
    },
  })
)

// Progressive: Load query subset immediately, full sync in background.
const collaborativeCollection = createCollection(
  queryCollectionOptions({ syncMode: 'progressive', /* ... */ })
)

Live Queries

Basic Query

import { useLiveQuery } from '@tanstack/react-db'
import { eq } from '@tanstack/db'

function TodoList() {
  const { data: todos } = useLiveQuery((query) =>
    query
      .from({ todos: todoCollection })
      .where(({ todos }) => eq(todos.completed, false))
  )

  return {todos.map(todo => {todo.text})}
}

Query Builder API

const { data } = useLiveQuery((q) =>
  q
    .from({ t: todoCollection })
    .where(({ t }) => eq(t.status, 'active'))
    .orderBy(({ t }) => t.createdAt, 'desc')
    .limit(10)
)

Joins

const { data } = useLiveQuery((q) =>
  q
    .from({ t: todoCollection })
    .innerJoin(
      { u: userCollection },
      ({ t, u }) => eq(t.userId, u.id)
    )
    .innerJoin(
      { p: projectCollection },
      ({ u, p }) => eq(u.projectId, p.id)
    )
    .where(({ p }) => eq(p.id, currentProject.id))
)

Filter Operators

import { eq, lt, and } from '@tanstack/db'

// Equality
eq(field, value)

// Less than
lt(field, value)

// AND
and(eq(product.category, 'electronics'), lt(product.price, 100))

With Ordering and Limits

const { data } = useLiveQuery((q) =>
  q
    .from({ product: productsCollection })
    .where(({ product }) =>
      and(eq(product.category, 'electronics'), lt(product.price, 100))
    )
    .orderBy(({ product }) => product.price, 'asc')
    .limit(10)
)

Optimistic Mutations

Insert

todoCollection.insert({
  id: uuid(),
  text: 'New todo',
  completed: false,
})
// Immediately: updates all live queries referencing this collection
// Background: calls onInsert handler to sync with server
// On failure: automatic rollback

No Manual Boilerplate

| Before (TanStack Query only) | After (TanStack DB) | |-------------------------------|---------------------| | Manual onMutate for optimistic state | Automatic | | Manual onError rollback logic | Automatic | | Per-mutation cache invalidation | All live queries update automatically |

Query-Driven Sync (On-Demand)

Live queries automatically generate optimized network requests:

// This live query...
useLiveQuery((q) =>
  q.from({ product: productsCollection })
    .where(({ product }) => and(eq(product.category, 'electronics'), lt(product.price, 100)))
    .orderBy(({ product }) => product.price, 'asc')
    .limit(10)
)

// ...automatically generates:
// GET /api/products?category=electronics&price_lt=100&sort=price:asc&limit=10

Predicate Mapping

queryFn: async (ctx) => {
  const { filters, sorts, limit } = parseLoadSubsetOptions(ctx.meta?.loadSubsetOptions)

  const params = new URLSearchParams()
  filters.forEach(({ field, operator, value }) => {
    if (operator === 'eq') params.set(field.join('.'), String(value))
    else if (operator === 'lt') params.set(`${field.join('.')}_lt`, String(value))
  })
  if (limit) params.set('limit', String(limit))

  return fetch(`/api/products?${params}`).then(r => r.json())
}

Performance

| Operation | Latency | |-----------|---------| | Single row update (100k sorted collection) | ~0.7 ms | | Subsequent queries (after sync) | 50k), progressive (collaborative)

1. Use joins instead of view-specific APIs - load normalized collections once
2. Let TanStack Query handle fetching - DB augments Query, doesn't replace it
3. Use parseLoadSubsetOptions to map live query predicates to API params
4. Rely on automatic optimistic updates - don't manually manage optimistic state
5. Use schemas for runtime validation and TypeScript inference
6. Leverage incremental computation - let the engine handle filtering vs manual .filter()

Common Pitfalls

• Creating collections inside components (should be module-level)
• Trying to replace TanStack Query entirely (DB builds on top of it)
• Using manual .filter() in render instead of live query where clauses
• Not providing getKey for proper normalization
• Forgetting mutation handlers (onInsert, onUpdate, onDelete) for server sync

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: tanstack-skills
• Source: tanstack-skills/tanstack-skills
• License: MIT

Install and usage instructions live in the source repository linked above.