api-baas-appwrite

Appwrite Patterns

Quick Guide: Use Appwrite as your open-source BaaS for authentication, structured data (TablesDB), file storage, serverless functions, and realtime subscriptions. Always initialize service classes from a shared Client instance, set permissions explicitly on every row (nothing is accessible by default), and use the server SDK (node-appwrite) with API key auth only on the backend.

<critical_requirements>

CRITICAL: Before Using This Skill

All code must follow project conventions in CLAUDE.md (kebab-case, named exports, import ordering, import type, named constants)

(You MUST set permissions explicitly on every row and file — Appwrite grants NO access by default, so omitting permissions makes data inaccessible)

(You MUST use node-appwrite with API key auth on the server and appwrite with session auth on the client — NEVER expose API keys in client bundles)

(You MUST always check for AppwriteException on every SDK call — Appwrite throws exceptions, it does NOT return { data, error } tuples)

(You MUST use ID.unique() for auto-generated IDs — passing a raw string creates a custom ID, not an auto-generated one)

(You MUST use the TablesDB API for new projects — the legacy Databases/collections/documents API is deprecated and receives only security patches)

</critical_requirements>

Auto-detection: Appwrite, appwrite, node-appwrite, TablesDB, Account, Databases, Permission, Role, ID.unique, Query.equal, createEmailPasswordSession, realtime.subscribe, Channel.files, Channel.tablesdb, APPWRITE_FUNCTION

When to use:

• Setting up an Appwrite client or server SDK with TypeScript
• Implementing authentication (email/password, OAuth, magic URL, phone OTP, anonymous sessions)
• Performing CRUD on TablesDB tables (rows, queries, permissions)
• Uploading and serving files from Storage buckets with access control
• Building serverless functions triggered by events, schedules, or HTTP
• Subscribing to realtime changes on rows, files, or account events
• Managing team memberships and role-based permissions

Key patterns covered:

• Client and server SDK initialization with typed service classes
• Auth flows: sign up, sign in, OAuth, magic URL, session management
• TablesDB operations with Query class for filtering, pagination, ordering
• Permission model: Permission.read(Role.any()), Role.user(), Role.team()
• Storage: upload, download, preview with image transforms, bucket permissions
• Serverless functions: export default async ({ req, res, log, error }) => {}
• Realtime subscriptions via Channel helpers and realtime.subscribe()

When NOT to use:

• Direct database connections or SQL queries (Appwrite is API-only, no raw SQL)
• Complex relational joins across many tables (Appwrite supports relationships but not arbitrary SQL joins)
• Applications requiring server-side realtime (Appwrite Realtime is client SDK only)

Detailed Resources:

• For decision frameworks and anti-patterns, see [reference.md](reference.md)

Client & Queries:

• [examples/core.md](examples/core.md) — Client setup, TablesDB CRUD, error handling, permissions

Authentication:

• [examples/auth.md](examples/auth.md) — Full auth flows, OAuth, magic URL, sessions, teams

Storage & Functions:

• [examples/storage.md](examples/storage.md) — File upload, download, previews, bucket permissions
• [examples/functions.md](examples/functions.md) — Serverless functions, triggers, SDK usage inside functions

Realtime:

• [examples/realtime.md](examples/realtime.md) — Channel subscriptions, event filtering, cleanup

<philosophy>

Philosophy

Appwrite is an open-source backend-as-a-service providing authentication, databases (TablesDB), file storage, serverless functions, and realtime — all through a unified SDK. It can be self-hosted or used via Appwrite Cloud.

Core principles:

1. Secure by default — Nothing is accessible without explicit permissions. Every row and file must have permissions set, or it is invisible to all users. This is the opposite of "open by default" — treat it as a mandatory step, not an afterthought.
2. Service class architecture — All SDK interactions go through service classes (Account, TablesDB, Storage, Functions, Teams, Realtime) instantiated from a shared Client. This pattern is consistent across client and server SDKs.
3. Two SDK split — appwrite (client) uses session-based auth for browsers. node-appwrite (server) uses API key auth for backends. They share the same API shape but different auth mechanisms. Never mix them.
4. Exceptions, not tuples — Appwrite throws AppwriteException on failure, not { data, error } tuples. Always wrap calls in try/catch.
5. TablesDB is the future — Appwrite 1.8 introduced TablesDB (tables/rows/columns) as the modern API. The legacy Databases API (collections/documents/attributes) still works but is deprecated. New features only land in TablesDB.
6. ID generation — Use ID.unique() for server-generated unique IDs. If you pass a string directly, it becomes a custom ID (which must be globally unique within the table).

When to use Appwrite:

• Open-source, self-hostable BaaS with full control over your data
• Projects needing auth, database, storage, and functions in one platform
• Teams wanting a self-hostable BaaS with no vendor lock-in
• Applications that benefit from granular document/row-level permissions

When NOT to use:

• Complex SQL-heavy applications requiring joins, views, and stored procedures
• Server-side realtime consumers (Appwrite Realtime is client SDK only — no server SDK support)
• Offline-first apps requiring local-first sync (Appwrite has no built-in offline sync)

</philosophy>

<patterns>

Core Patterns

Pattern 1: Client SDK Setup (Browser)

Initialize the Appwrite client with endpoint and project ID. All service classes share a single Client instance.

// lib/appwrite.ts
import { Client, Account, TablesDB, Storage, Realtime } from "appwrite";

const APPWRITE_ENDPOINT = process.env.APPWRITE_ENDPOINT!;
const APPWRITE_PROJECT_ID = process.env.APPWRITE_PROJECT_ID!;

const client = new Client()
  .setEndpoint(APPWRITE_ENDPOINT)
  .setProject(APPWRITE_PROJECT_ID);

export const account = new Account(client);
export const tablesDB = new TablesDB(client);
export const storage = new Storage(client);
export const realtime = new Realtime(client);

Why good: Single client instance shared across services, named constants for config, named exports for each service, environment variables keep config out of code

// BAD: Hardcoded config, recreating clients
import { Client, Account } from "appwrite";

const account = new Account(
  new Client()
    .setEndpoint("https://cloud.appwrite.io/v1") // Hardcoded
    .setProject("abc123"), // Hardcoded project ID
);

Why bad: Hardcoded endpoint and project ID, new Client per service wastes resources, no shared config

Pattern 2: Server SDK Setup (Node.js)

The server SDK uses API key authentication — never expose API keys in client code.

// lib/appwrite-server.ts
import { Client, TablesDB, Users, Storage } from "node-appwrite";

const APPWRITE_ENDPOINT = process.env.APPWRITE_ENDPOINT!;
const APPWRITE_PROJECT_ID = process.env.APPWRITE_PROJECT_ID!;
const APPWRITE_API_KEY = process.env.APPWRITE_API_KEY!;

const client = new Client()
  .setEndpoint(APPWRITE_ENDPOINT)
  .setProject(APPWRITE_PROJECT_ID)
  .setKey(APPWRITE_API_KEY);

export const tablesDB = new TablesDB(client);
export const users = new Users(client);
export const storage = new Storage(client);

Why good: .setKey() for API key auth (server only), Users service for admin user management (not available in client SDK), named constants

When to use: API routes, serverless functions, admin scripts, webhooks. NEVER import this module from clien