Cloudflare Platform Products

Cloudflare Platform Products

Purpose

This skill provides guidance on Cloudflare's platform products and services that integrate with Workers. It covers storage options (KV, R2, D1), coordination services (Durable Objects, Queues), data services (Vectorize, Hyperdrive), and specialized services (Analytics Engine, Browser Rendering). Use this skill when choosing between platform products, designing system architecture, or integrating multiple Cloudflare services.

Platform Products Overview

Cloudflare offers a comprehensive suite of platform products designed to work seamlessly with Workers:

See references/platform-products-matrix.md for detailed comparison and selection guide.

Storage Services

KV (Key-Value Storage)

Best for: Static content, configuration, caching, read-heavy workloads

Characteristics:

• Eventually consistent (writes propagate in ~60 seconds globally)
• Optimized for reads (not writes)
• 25 MB max value size
• Global replication included
• Metadata support

When to use:

• Caching API responses
• Storing static assets
• Configuration data
• Session data (with expiration)
• Read-heavy data with infrequent writes

When NOT to use:

• Frequently changing data
• Strong consistency requirements
• Large objects (> 25 MB)
• Complex queries

Example use cases:

// Cache API responses
await env.CACHE.put(`api:users:${id}`, JSON.stringify(user), {
  expirationTtl: 3600
});

// Store configuration
await env.CONFIG.put('feature_flags', JSON.stringify(flags));

// Session storage
await env.SESSIONS.put(sessionId, userData, {
  expirationTtl: 86400 // 24 hours
});

D1 (SQLite Database)

Best for: Relational data, complex queries, transactional workloads

Characteristics:

• SQLite database
• Strong consistency
• ACID transactions
• SQL query support
• Migrations support
• 25 MB database size (beta limit)

When to use:

• Structured relational data
• Complex queries with JOINs
• Transactional operations
• Data with relationships
• Migrations-driven schema

When NOT to use:

• Large datasets (> 25 MB in beta)
• Very high write throughput
• Unstructured data
• Simple key-value lookups (use KV instead)

Example use cases:

// User management
await env.DB.prepare(
  'SELECT * FROM users WHERE email = ? AND active = 1'
).bind(email).first();

// Transactions (via batch)
await env.DB.batch([
  env.DB.prepare('INSERT INTO orders (user_id, total) VALUES (?, ?)').bind(userId, total),
  env.DB.prepare('UPDATE users SET last_order = ? WHERE id = ?').bind(Date.now(), userId)
]);

// Complex queries
await env.DB.prepare(`
  SELECT orders.*, users.email
  FROM orders
  JOIN users ON orders.user_id = users.id
  WHERE orders.status = ?
`).bind('pending').all();

R2 (Object Storage)

Best for: Large files, user uploads, backups, media storage

Characteristics:

• S3-compatible API
• No size limits per object
• Zero egress fees (from Workers)
• Custom metadata
• Streaming support

When to use:

• Large files (> 25 MB)
• User-uploaded content
• Media files (images, videos)
• Backups and archives
• Static website hosting

When NOT to use:

• Small values (< 1 KB, use KV)
• Frequently updated small data
• Requires low-latency for small reads (KV is faster)

Example use cases:

// Store user upload
await env.UPLOADS.put(`users/${userId}/avatar.jpg`, imageData, {
  httpMetadata: {
    contentType: 'image/jpeg'
  },
  customMetadata: {
    uploadedBy: userId,
    uploadedAt: Date.now().toString()
  }
});

// Stream large file
const object = await env.MEDIA.get('videos/large-video.mp4');
return new Response(object.body);

// Store backup
await env.BACKUPS.put(`db-backup-${Date.now()}.sql`, backupData);

See references/storage-options-guide.md for detailed storage selection criteria.

Coordination Services

Durable Objects

Best for: Coordination, real-time collaboration, WebSockets, rate limiting

Characteristics:

• Strong consistency
• Stateful instances
• Single-threaded execution per object
• Persistent storage
• WebSocket support
• Global coordination

When to use:

• Real-time collaboration (chat, docs)
• Rate limiting and quotas
• Coordination between distributed requests
• Persistent WebSocket connections
• Stateful game servers
• Sequential processing requirements

When NOT to use:

• Simple stateless operations
• Read-heavy workloads (use KV)
• Pure data storage (use D1/R2)
• High-throughput parallel processing

Example use cases:

// Rate limiting
export class RateLimiter {
  constructor(state, env) {
    this.state = state;
  }

  async fetch(request) {
    const count = await this.state.storage.get('count') || 0;
    const limit = 100;

    if (count >= limit) {
      return new Response('Rate limit exceeded', { status: 429 });
    }

    await this.state.storage.put('count', count + 1);
    return new Response('OK');
  }
}

// Chat room coordination
export class ChatRoom {
  constructor(state, env) {
    this.state = state;
    this.sessions = [];
  }

  async fetch(request) {
    const pair = new WebSocketPair();
    this.sessions.push(pair[1]);

    pair[1].accept();
    pair[1].addEventListener('message', event => {
      // Broadcast to all sessions
      this.sessions.forEach(session => {
        session.send(event.data);
      });
    });

    return new Response(null, { status: 101, webSocket: pair[0] });
  }
}

Queues

Best for: Asynchronous processing, background jobs, event-driven workflows

Characteristics:

• At-least-once delivery
• Message batching
• Dead letter queues
• Automatic retries
• Guaranteed ordering per message

When to use:

• Background processing
• Webhook handling
• Email sending
• Data pipeline stages
• Decoupling services
• Batch processing

When NOT to use:

• Synchronous request-response
• Exactly-once delivery required (handle idempotency in consumer)
• Real-time requirements (use Durable Objects + WebSockets)

Example use cases:

// Producer: Queue email sending
await env.EMAIL_QUEUE.send({
  to: user.email,
  subject: 'Welcome',
  body: 'Welcome to our service!'
});

// Consumer: Process emails in batches
export default {
  async queue(batch, env) {
    for (const message of batch.messages) {
      const { to, subject, body } = message.body;

      try {
        await sendEmail(to, subject, body, env);
        message.ack();
      } catch (error) {
        console.error('Email failed:', error);
        message.retry();
      }
    }
  }
};

Data Services

Vectorize

Best for: Semantic search, RAG, recommendations, similarity matching

Characteristics:

• Vector similarity search
• Configurable dimensions (matching embedding model)
• Metadata storage
• Batch operations
• Cosine/Euclidean/Dot product metrics

When to use:

• Semantic search
• RAG (Retrieval Augmented Generation)
• Recommendation systems
• Image similarity
• Duplicate detection
• Content clustering

When NOT to use:

• Exact text search (use D1 with LIKE or full-text search)
• Simple key-value lookup (use KV)
• Small datasets that fit in memory

Example use cases:

// Generate and store embeddings
const embeddings = await env.AI.run('@cf/baai/bge-base-en-v1.5', {
  text: [document.text]
});

await env.VECTOR_INDEX.insert([{
  id: document.id,
  values: embeddings.data[0],
  metadata: { text: document.text, title: document.title }
}]);

// Semantic search
const queryEmbedding = await env.AI.run('@cf/baai/bge-base-en-v1.5', {
  text: [userQuestion]
});

const results = await env.VECTOR_INDEX.query(queryEmbedding.data[0], {
  topK: 5
});

// Use results for RAG
const context = results.matches.map(m => m.metadata.text).join('\n');

Hyperdrive

Best for: Postgres connection pooling, reducing database latency

Characteristics:

• Connection pooling for Postgres
• Reduces connection overhead
• Regional caching
• Automatic connection management

When to use:

• Connecting to external Postgres databases
• High connection churn
• Reducing latency to databases
• Traditional database migration to Workers

When NOT to use:

• SQLite databases (use D1)
• Non-Postgres databases
• When D1 meets requirements

Example use cases:

// Connect to Postgres via Hyperdrive
const client = env.HYPERDRIVE.connect();

const result = await client.query(
  'SELECT * FROM users WHERE id = $1',
  [userId]
);

// Connection automatically pooled and managed

Analytics and Observability

Analytics Engine

Best for: Custom metrics, high-cardinality analytics, time-series data

Characteristics:

• Write-optimized time-series database
• SQL query support
• High cardinality support
• Automatic aggregation

When to use:

• Application metrics
• User analytics
• Performance monitoring
• Business intelligence
• Custom event tracking

When NOT to use:

• Real-time queries (data available after ~1 minute)
• Transactional data (use D1)
• Simple counters (use Durable Objects)

Example use cases:

// Write analytics events
await env.ANALYTICS.writeDataPoint({
  blobs: ['api_call', request.url, request.method],
  doubles: [responseTime, statusCode],
  indexes: [userId]
});

// Query via GraphQL API or Dashboard

Specialized Services

Browser Rendering

Best for: Web scraping, PDF generation, screenshots, browser automation

Characteristics:

• Headless Chrome browser
• Puppeteer API compatibility
• Full browser environment
• JavaScript execution

When to use:

• Taking screenshots
• Generating PDFs
• Web scraping dynamic sites
• Testing web pages
• Browser automation

Example use cases:

// Take screenshot
const browser = await puppeteer.launch(env.BROWSER);
const page = await browser.newPage();
await page.goto('https://example.com');
const screenshot = await page.screenshot();
await browser.close();

return new Response(screenshot, {
  headers: { 'Content-Type': 'image/png' }
});

Email Routing

Best for: Custom email handling, email forwarding, email processing

Characteristics:

• Programmable email handling
• Email parsing
• Forward or drop emails
• Integration with Workers

When to use:

• Custom email routing logic
• Email processing pipelines
• Spam filtering
• Email-triggered workflows

Service Selection Guide

Storage Decision Tree

Need to store data → What kind?

1. Large files (> 25 MB) or media? → Use R2
2. Relational data with complex queries? → Use D1
3. Key-value, read-heavy, cache? → Use KV
4. Vector embeddings for search? → Use Vectorize

Processing Decision Tree

Need to process requests → What pattern?

1. Real-time coordination, WebSockets? → Use Durable Objects
2. Async background jobs? → Use Queues
3. Stateless request processing? → Use Workers (no special binding)
4. Rate limiting per user/IP? → Use Durable Objects

Integration Patterns

Common multi-product patterns:

1. RAG Application:

   • Workers AI (embeddings)
   • Vectorize (vector storage)
   • D1 (original text storage)
   • Workers AI (text generation)

2. E-commerce Platform:

   • D1 (product catalog, orders)
   • R2 (product images)
   • KV (session data, cache)
   • Queues (order processing)
   • Durable Objects (inventory management)

3. Content Platform:

   • R2 (media files)
   • D1 (metadata, users)
   • KV (CDN cache)
   • Analytics Engine (usage metrics)

4. Real-time Collaboration:

   • Durable Objects (room coordination)
   • D1 (persistent data)
   • R2 (file attachments)
   • Queues (notifications)

See examples/multi-product-architecture.js for complete integration examples.

Cost Optimization

General Principles

1. Right-size storage: Use KV for small data, R2 for large files
2. Cache effectively: Use KV to cache D1 queries or API responses
3. Batch operations: Use Queue batching, D1 batch(), Vectorize bulk inserts
4. Use Workers AI + Vectorize: No egress costs between services
5. Leverage free tiers: Most products have generous free tiers

Free Tier Limits

• KV: 100k reads/day, 1k writes/day, 1 GB storage
• D1: 5 million rows read, 100k rows written
• R2: 10 GB storage, 1 million Class A operations
• Queues: 1 million operations/month
• Vectorize: 30 million queried vectors/month
• Workers AI: Limited free inference requests

See references/pricing-optimization.md for detailed cost optimization strategies.

Migration Patterns

From Traditional Database to D1

// Before: External Postgres via Hyperdrive
const result = await env.HYPERDRIVE.query('SELECT * FROM users WHERE id = ?', [id]);

// After: D1 (if dataset fits in 25 MB)
const result = await env.DB.prepare('SELECT * FROM users WHERE id = ?').bind(id).first();

From S3 to R2

// R2 is S3-compatible, minimal code changes needed
// Before: aws-sdk with S3
// After: R2 binding (native integration)

await env.MY_BUCKET.put('key', data);
const object = await env.MY_BUCKET.get('key');

From Redis to KV or Durable Objects

// Simple cache: KV
await env.CACHE.put('key', 'value', { expirationTtl: 3600 });

// Stateful/counters: Durable Objects
const id = env.COUNTER.idFromName('global');
const counter = env.COUNTER.get(id);
await counter.fetch(request);

Best Practices

Storage Selection

• Hot data: KV (frequently accessed, rarely changed)
• Warm data: D1 (structured, occasional queries)
• Cold data: R2 (archival, backups)

Consistency Requirements

• Strong consistency needed: D1, Durable Objects
• Eventually consistent OK: KV, Vectorize
• At-least-once delivery: Queues

Performance Optimization

• Cache D1 queries in KV for read-heavy workloads
• Use Vectorize batch inserts for efficiency
• Leverage Durable Objects for coordination, not storage
• Stream large R2 objects instead of buffering

Security

• Use Cloudflare Access for private buckets/databases
• Validate all user input before storing
• Use signed URLs for temporary R2 access
• Implement rate limiting with Durable Objects
• Encrypt sensitive data before storing in KV/R2

Additional Resources

Reference Files

For detailed information, consult:

• references/platform-products-matrix.md - Complete product comparison and selection criteria
• references/storage-options-guide.md - Deep dive on KV vs D1 vs R2 decisions
• references/pricing-optimization.md - Cost optimization strategies

Example Files

Working examples in examples/:

• multi-product-architecture.js - Integrating multiple platform products

Documentation Links

For the latest platform documentation:

• Platform overview: https://developers.cloudflare.com/products/
• KV: https://developers.cloudflare.com/kv/
• D1: https://developers.cloudflare.com/d1/
• R2: https://developers.cloudflare.com/r2/
• Durable Objects: https://developers.cloudflare.com/durable-objects/
• Queues: https://developers.cloudflare.com/queues/
• Vectorize: https://developers.cloudflare.com/vectorize/

Use the cloudflare-docs-specialist agent to search documentation and fetch the latest platform information.