AI & Vectorize Bindings

Jack cloud projects can use Cloudflare's AI and Vectorize services for machine learning inference and vector search.

Quick Start

Add bindings to your wrangler.jsonc:

{
  "ai": {
    "binding": "AI"
  },
  "vectorize": [
    {
      "binding": "VECTORS",
      "index_name": "my-vectors",
      "preset": "cloudflare"
    }
  ]
}

Deploy and jack cloud automatically provisions the resources:

jack ship

AI Binding

The AI binding gives you access to Cloudflare's AI models for inference.

Configuration

{
  "ai": {
    "binding": "AI"
  }
}

Usage

interface Env {
  AI: Ai;
}
 
export default {
  async fetch(request: Request, env: Env) {
    // Text generation
    const response = await env.AI.run("@cf/meta/llama-3-8b-instruct", {
      prompt: "What is the capital of France?",
    });
 
    // Text embeddings
    const embeddings = await env.AI.run("@cf/baai/bge-base-en-v1.5", {
      text: "Hello world",
    });
 
    // Image classification
    const result = await env.AI.run("@cf/microsoft/resnet-50", {
      image: await request.arrayBuffer(),
    });
 
    return Response.json(response);
  },
};

Available Models

Common models include:

Model	Use Case
`@cf/meta/llama-3-8b-instruct`	Text generation, chat
`@cf/baai/bge-base-en-v1.5`	Text embeddings (768 dims)
`@cf/baai/bge-small-en-v1.5`	Smaller embeddings (384 dims)
`@cf/microsoft/resnet-50`	Image classification
`@cf/openai/whisper`	Speech to text

See Cloudflare AI Models for the full list.

Vectorize Binding

Vectorize is a vector database for semantic search, RAG, and similarity matching.

Configuration

{
  "vectorize": [
    {
      "binding": "VECTORS",
      "index_name": "my-index",
      "preset": "cloudflare"
    }
  ]
}

Presets:

Preset	Dimensions	Metric	Best For
`cloudflare`	768	cosine	BGE-base embeddings
`cloudflare-small`	384	cosine	BGE-small embeddings
`cloudflare-large`	1024	cosine	Larger models

Or specify dimensions manually:

{
  "vectorize": [
    {
      "binding": "VECTORS",
      "index_name": "custom-index",
      "dimensions": 1536,
      "metric": "cosine"
    }
  ]
}

Usage

interface Env {
  VECTORS: VectorizeIndex;
}
 
export default {
  async fetch(request: Request, env: Env) {
    // Insert vectors
    await env.VECTORS.insert([
      {
        id: "doc-1",
        values: [0.1, 0.2, ...], // 768 floats for cloudflare preset
        metadata: { title: "My Document" },
      },
    ]);
 
    // Query similar vectors
    const results = await env.VECTORS.query([0.1, 0.2, ...], {
      topK: 5,
      returnMetadata: "all",
    });
 
    // results.matches = [{ id, score, metadata }, ...]
    return Response.json(results);
  },
};

Combining AI + Vectorize

A common pattern is using AI for embeddings and Vectorize for search:

interface Env {
  AI: Ai;
  VECTORS: VectorizeIndex;
}
 
// Index a document
async function indexDocument(env: Env, id: string, content: string) {
  // Generate embedding
  const response = await env.AI.run("@cf/baai/bge-base-en-v1.5", {
    text: content,
  });
  const embedding = response.data[0];
 
  // Store in Vectorize
  await env.VECTORS.insert([
    {
      id,
      values: embedding,
      metadata: { preview: content.slice(0, 100) },
    },
  ]);
}
 
// Search documents
async function search(env: Env, query: string) {
  // Generate query embedding
  const response = await env.AI.run("@cf/baai/bge-base-en-v1.5", {
    text: query,
  });
  const embedding = response.data[0];
 
  // Find similar documents
  const results = await env.VECTORS.query(embedding, {
    topK: 5,
    returnMetadata: "all",
  });
 
  return results.matches;
}

Templates

The semantic-search template includes a working AI + Vectorize setup:

jack new my-search -t semantic-search

This gives you:

Document indexing API
Semantic search API
Web UI for testing
D1 database for full content storage

Jack Cloud Provisioning

When you deploy with jack ship, jack cloud:

Detects AI and Vectorize bindings in your config
Provisions a Vectorize index (named jack-{projectId}-{binding})
Binds the resources to your worker
Handles all Cloudflare API setup automatically

You don't need a Cloudflare account or API tokens - jack cloud manages everything.

Local Development

For local development with wrangler dev, you need a Cloudflare account with AI and Vectorize enabled:

# Create index for local dev
wrangler vectorize create my-vectors --preset cloudflare
 
# Run locally
wrangler dev

Pricing

Jack cloud includes generous free tiers:

AI: Free tier included with Cloudflare Workers
Vectorize: 30M queried dimensions/month free, 5M stored dimensions free

See Cloudflare pricing for details.