Algolia Data Handling

Algolia Data Handling

Overview

Algolia stores your records in their cloud. You control what data goes in (via saveObjects), what comes back (via attributesToRetrieve), and what users can search (via searchableAttributes). For privacy compliance, you must filter PII before indexing and implement deletion workflows.

Prerequisites

• A documented data classification for the source records and approval to index only the searchable fields.
• An Algolia Admin key held by the backend indexing job, never the browser.
• A deletion and retention owner for data-subject requests before production records are copied to Algolia.

Data Flow: Source → Algolia → User

Source Database        Transform           Algolia Index           Search Response
┌──────────┐     ┌──────────────┐     ┌──────────────────┐     ┌──────────────┐
│ Full user │     │ Strip PII    │     │ Searchable       │     │ Retrieved    │
│ record    │ ──▶ │ Truncate     │ ──▶ │ fields only      │ ──▶ │ fields only  │
│ (all cols)│     │ Normalize    │     │ + ranking data   │     │ (UI needs)   │
└──────────┘     └──────────────┘     └──────────────────┘     └──────────────┘

Instructions

Examples

The transformations below demonstrate an allowlist-first record shape, explicit PII removal, deletion handling, and bounded retention. Apply the same policy to every source connector rather than relying on dashboard settings alone.

Step 1: Transform Records Before Indexing

import { algoliasearch } from 'algoliasearch';

const client = algoliasearch(process.env.ALGOLIA_APP_ID!, process.env.ALGOLIA_ADMIN_KEY!);

// Define what goes into Algolia — NOT everything from your DB
interface AlgoliaProduct {
  objectID: string;
  name: string;
  description: string;     // Truncated, plain text
  category: string;
  brand: string;
  price: number;
  in_stock: boolean;
  image_url: string;
  rating: number;
  _tags: string[];
}

function transformForAlgolia(dbRecord: any): AlgoliaProduct {
  return {
    objectID: dbRecord.id,
    name: dbRecord.name,
    description: stripHtml(dbRecord.description).substring(0, 5000),
    category: dbRecord.category?.name || 'uncategorized',
    brand: dbRecord.brand?.name || '',
    price: dbRecord.price_cents / 100,
    in_stock: dbRecord.inventory_count > 0,
    image_url: dbRecord.images?.[0]?.url || '',
    rating: dbRecord.avg_rating || 0,
    _tags: buildTags(dbRecord),
  };
}

// Strip HTML tags for clean search text
function stripHtml(html: string): string {
  return html?.replace(/<[^>]*>/g, ' ').replace(/\s+/g, ' ').trim() || '';
}

function buildTags(record: any): string[] {
  const tags: string[] = [];
  if (record.is_featured) tags.push('featured');
  if (record.is_new) tags.push('new-arrival');
  if (record.discount_percent > 0) tags.push('on-sale');
  return tags;
}

Step 2: PII Detection and Filtering

// NEVER index PII unless absolutely necessary for search
const PII_FIELDS = ['email', 'phone', 'ssn', 'address', 'credit_card', 'password', 'api_key'];

function stripPII(record: Record<string, any>): Record<string, any> {
  const clean = { ...record };
  for (const field of PII_FIELDS) {
    delete clean[field];
  }
  return clean;
}

// If you MUST index user-facing names (e.g., author names in articles)
// Use unretrievableAttributes so they're searchable but never returned
await client.setSettings({
  indexName: 'articles',
  indexSettings: {
    searchableAttributes: ['title', 'author_name', 'content'],
    unretrievableAttributes: ['author_name'],  // Searchable but never in response
    attributesToRetrieve: ['title', 'excerpt', 'url', 'published_at'],
  },
});

Step 3: Algolia-Side Data Access Control

// Use secured API keys to filter what each user can see
function generateUserKey(userId: string, tenantId: string) {
  return client.generateSecuredApiKey({
    parentApiKey: process.env.ALGOLIA_SEARCH_KEY!,
    restrictions: {
      filters: `tenant_id:${tenantId} AND (visibility:public OR created_by:${userId})`,
      validUntil: Math.floor(Date.now() / 1000) + 3600,
    },
  });
}

// User can only search records where:
// - tenant_id matches their org AND
// - visibility is public OR they created it

Step 4: GDPR Right to Deletion

// When a user requests data deletion:
async function deleteUserData(userId: string) {
  const results: Record<string, string> = {};

  // 1. Delete user's records from all indices
  for (const indexName of ['products', 'reviews', 'wishlists']) {
    try {
      await client.deleteBy({
        indexName,
        deleteByParams: { filters: `created_by:${userId}` },
      });
      results[indexName] = 'deleted';
    } catch (e) {
      results[indexName] = `failed: ${e}`;
    }
  }

  // 2. Delete Insights/Analytics data for this user
  // Algolia retains events for 90 days by default
  // Use the Insights API to request user data deletion
  await client.deleteUserToken({ userToken: userId });

  // 3. Log the deletion for compliance audit
  console.log({
    event: 'gdpr.deletion',
    userId,
    timestamp: new Date().toISOString(),
    results,
  });

  return results;
}

Step 5: Data Subject Access Request (DSAR)

// Export all data associated with a user
async function exportUserData(userId: string) {
  const exportData: Record<string, any[]> = {};

  for (const indexName of ['products', 'reviews', 'wishlists']) {
    const records: any[] = [];
    let cursor: string | undefined;

    // Browse all records matching the user
    do {
      const result = await client.browse({
        indexName,
        browseParams: {
          filters: `created_by:${userId}`,
          hitsPerPage: 1000,
          cursor,
        },
      });
      records.push(...result.hits);
      cursor = result.cursor;
    } while (cursor);

    exportData[indexName] = records;
  }

  return {
    exportedAt: new Date().toISOString(),
    userId,
    data: exportData,
  };
}

Step 6: Data Retention and Cleanup

// Scheduled job: delete old records past retention period
async function enforceRetention(indexName: string, retentionDays: number) {
  const cutoffTimestamp = Math.floor(
    (Date.now() - retentionDays * 24 * 60 * 60 * 1000) / 1000
  );

  await client.deleteBy({
    indexName,
    deleteByParams: {
      filters: `created_at_timestamp < ${cutoffTimestamp}`,
    },
  });

  console.log(`Deleted records older than ${retentionDays} days from ${indexName}`);
}

// Run daily: enforceRetention('activity_logs', 90);

Data Classification for Algolia

Output

The index contains only the approved, normalized record fields; data-subject deletion and retention jobs can remove those records deterministically. The resulting search response exposes only fields the UI is permitted to retrieve.

Error Handling

Resources

• Algolia Privacy & GDPR
• deleteBy Reference
• browse Reference
• Insights User Deletion

Next Steps

For enterprise access control, see algolia-enterprise-rbac.