Algolia sdk patterns
'Apply production-ready algoliasearch v5 patterns: singleton client, typed search,From its SKILL.md
npx -y skills add jeremylongshore/claude-code-plugins-plus-skills --skill algolia-sdk-patternsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
6.6 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it
Algolia SDK Patterns
Overview
Production-ready patterns for algoliasearch v5. Key architectural change from v4: all methods live on the client directly — no more client.initIndex(). Index name is passed as a parameter to every call.
Prerequisites
algoliasearchv5+ installed- Completed
algolia-install-authsetup - TypeScript project (patterns work in JS too, you just lose type safety)
Instructions
Pattern 1: Typed Singleton Client
// src/algolia/client.ts
import { algoliasearch, type Algoliasearch } from 'algoliasearch';
let _client: Algoliasearch | null = null;
export function getClient(): Algoliasearch {
if (!_client) {
const appId = process.env.ALGOLIA_APP_ID;
const apiKey = process.env.ALGOLIA_ADMIN_KEY;
if (!appId || !apiKey) {
throw new Error(
'ALGOLIA_APP_ID and ALGOLIA_ADMIN_KEY must be set. '
+ 'Get them from dashboard.algolia.com > Settings > API Keys'
);
}
_client = algoliasearch(appId, apiKey);
}
return _client;
}
// For testing: reset singleton
export function resetClient(): void {
_client = null;
}
Pattern 2: Typed Search Results
// src/algolia/types.ts
// Define your record shape — extends Algolia's Hit type
interface Product {
objectID: string;
name: string;
category: string;
price: number;
description: string;
image_url: string;
}
// src/algolia/search.ts
import { getClient } from './client';
export async function searchProducts(
query: string,
options?: {
filters?: string;
facetFilters?: string[][];
hitsPerPage?: number;
page?: number;
}
) {
const client = getClient();
const { hits, nbHits, nbPages, page } = await client.searchSingleIndex<Product>({
indexName: 'products',
searchParams: {
query,
filters: options?.filters,
facetFilters: options?.facetFilters,
hitsPerPage: options?.hitsPerPage ?? 20,
page: options?.page ?? 0,
attributesToRetrieve: ['name', 'category', 'price', 'image_url'],
attributesToHighlight: ['name', 'description'],
},
});
return { hits, totalHits: nbHits, totalPages: nbPages, currentPage: page };
}
// Usage: const { hits } = await searchProducts('laptop', { filters: 'price < 1000' });
Pattern 3: Error Handling with Algolia Error Types
// src/algolia/errors.ts
import { ApiError } from 'algoliasearch';
export async function safeAlgoliaCall<T>(
operation: string,
fn: () => Promise<T>
): Promise<{ data: T | null; error: string | null }> {
try {
const data = await fn();
return { data, error: null };
} catch (err) {
if (err instanceof ApiError) {
// ApiError has status and message from Algolia API
const msg = `Algolia ${operation} failed [${err.status}]: ${err.message}`;
console.error(msg);
// Specific handling for common codes
if (err.status === 429) {
console.warn('Rate limited — reduce request frequency or contact Algolia');
} else if (err.status === 404) {
console.warn('Index or object not found — verify index name');
}
return { data: null, error: msg };
}
// Non-Algolia error (network, etc.)
const msg = err instanceof Error ? err.message : 'Unknown error';
console.error(`${operation} error: ${msg}`);
return { data: null, error: msg };
}
}
// Usage:
// const { data, error } = await safeAlgoliaCall('search', () =>
// client.searchSingleIndex({ indexName: 'products', searchParams: { query: 'foo' } })
// );
Pattern 4: Batch Operations
// src/algolia/batch.ts
import { getClient } from './client';
// saveObjects handles batching internally — send up to 1000 objects per call
export async function bulkIndex(indexName: string, records: Record<string, any>[]) {
const client = getClient();
const BATCH_SIZE = 1000;
for (let i = 0; i < records.length; i += BATCH_SIZE) {
const batch = records.slice(i, i + BATCH_SIZE);
const { taskID } = await client.saveObjects({
indexName,
objects: batch,
});
await client.waitForTask({ indexName, taskID });
console.log(`Indexed ${Math.min(i + BATCH_SIZE, records.length)}/${records.length}`);
}
}
// Partial update — only send changed fields
export async function updateFields(
indexName: string,
objectID: string,
fields: Record<string, any>
) {
const client = getClient();
return client.partialUpdateObject({
indexName,
objectID,
attributesToUpdate: fields,
});
}
Pattern 5: Multi-Tenant Client Factory
// src/algolia/multi-tenant.ts
import { algoliasearch, type Algoliasearch } from 'algoliasearch';
const tenantClients = new Map<string, Algoliasearch>();
export function getClientForTenant(tenantId: string): Algoliasearch {
if (!tenantClients.has(tenantId)) {
// Each tenant might have their own Algolia app, or use index prefixes
const appId = process.env[`ALGOLIA_APP_ID_${tenantId.toUpperCase()}`]
|| process.env.ALGOLIA_APP_ID!;
const apiKey = process.env[`ALGOLIA_ADMIN_KEY_${tenantId.toUpperCase()}`]
|| process.env.ALGOLIA_ADMIN_KEY!;
tenantClients.set(tenantId, algoliasearch(appId, apiKey));
}
return tenantClients.get(tenantId)!;
}
// Or use a single app with index prefixing
export function tenantIndex(tenantId: string, base: string): string {
return `${tenantId}_${base}`; // "acme_products"
}
Error Handling
| Pattern | Use Case | Benefit |
|---|---|---|
safeAlgoliaCall wrapper | All API calls | Prevents uncaught exceptions, structured error info |
ApiError check | Distinguishing API vs network errors | Targeted retry/recovery logic |
waitForTask | After every write operation | Ensures reads see latest data |
| Batch chunking | Large datasets | Avoids record-too-big and timeout errors |
Resources
- algoliasearch v5 Methods
- v4 to v5 Migration Guide
- API Error Codes
Next Steps
Apply patterns in algolia-core-workflow-a for search implementation.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.