AI-Powered Features¶

Overview¶

Centrali integrates AI capabilities throughout the platform to help you work smarter with your data. From generating schemas to detecting anomalies, AI features reduce manual work and catch issues before they become problems.

Feature Summary¶

AI Schema Generator¶

Describe your data structure in plain English and get a complete schema with fields, types, and validation rules.

How It Works¶

1. Navigate to Structures in the Console
2. Click Create Structure
3. Select AI Generate option
4. Describe your data in natural language
5. Review and customize the suggested schema
6. Save your structure

Example¶

Input:

"I need a structure for tracking customer orders. Each order has a customer email, order date, list of items with name and price, total amount, and status that can be pending, processing, shipped, or delivered."

Generated Schema:

{
  "type": "object",
  "properties": {
    "customerEmail": { "type": "string", "format": "email" },
    "orderDate": { "type": "string", "format": "date-time" },
    "items": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "name": { "type": "string" },
          "price": { "type": "number", "minimum": 0 }
        }
      }
    },
    "totalAmount": { "type": "number", "minimum": 0 },
    "status": {
      "type": "string",
      "enum": ["pending", "processing", "shipped", "delivered"]
    }
  }
}

Natural Language Search¶

Search your records using conversational queries instead of complex filter syntax.

How It Works¶

1. Open any structure's records view
2. Click the AI Search button or press /
3. Type your query in natural language
4. The AI translates it to filters and returns results

Example Queries¶

SDK Usage¶

// Natural language search via SDK
const results = await centrali.records.nlSearch('orders', {
  query: 'orders from last week with status shipped'
});

console.log('Found:', results.data.length, 'records');
console.log('Translated filter:', results.meta.translatedFilter);

Natural Language Record Creation¶

Create records by describing them instead of filling out forms field by field.

How It Works¶

1. Open a structure's records view
2. Click Create Record
3. Select AI Create mode
4. Describe the record you want to create
5. Review the extracted fields
6. Confirm and save

Example¶

Input:

"Add a new customer John Smith with email john.smith@example.com, he signed up today and is on the pro plan"

Extracted Record:

{
  "name": "John Smith",
  "email": "john.smith@example.com",
  "signupDate": "2024-01-15T00:00:00Z",
  "plan": "pro"
}

Validation¶

The AI validates extracted data against your schema. If a required field is missing or a value doesn't match the expected type, you'll be prompted to correct it before saving.

Compute AI Assistant¶

Generate compute function code by describing what you want to accomplish.

How It Works¶

1. Create or edit a compute function
2. Click AI Generate in the code editor
3. Describe what the function should do
4. Review and customize the generated code
5. Test and deploy

Example¶

Input:

"When an order is created, send an email to the customer with their order details and estimated delivery date"

Generated Code:

export async function handler(event, context) {
  const { api, handlebars } = context;
  const order = event.data;

  // Get customer details
  const customer = await api.records.get('customers', order.customerId);

  // Calculate estimated delivery (5 business days)
  const deliveryDate = new Date();
  deliveryDate.setDate(deliveryDate.getDate() + 7);

  // Render email template
  const emailBody = handlebars.compile(`
    Hi {{customer.name}},

    Thank you for your order #{{order.id}}!

    Order Summary:
    {{#each order.items}}
    - {{this.name}}: ${{this.price}}
    {{/each}}

    Total: ${{order.totalAmount}}
    Estimated Delivery: {{formatDate deliveryDate "MMMM D, YYYY"}}
  `)({ customer: customer.data, order, deliveryDate });

  // Send email
  await api.notifications.sendEmail({
    to: customer.data.email,
    subject: `Order Confirmation #${order.id}`,
    body: emailBody
  });

  return { success: true };
}

Capabilities¶

The AI assistant understands:

• Centrali's compute runtime APIs (api.records, api.notifications, etc.)
• Event types and payloads
• Common patterns (CRUD operations, notifications, webhooks)
• External API integrations (via fetch)

AI Data Validation¶

Automatically detect and fix data quality issues in your records.

Read the full AI Data Validation guide →

Quick Overview¶

• Real-time validation as records are created or updated
• Batch scanning for existing data
• Auto-correct mode for high-confidence fixes
• Issue types: Format errors, typos, duplicates, semantic issues

Anomaly Detection & Insights¶

AI-powered detection of unusual patterns, statistical outliers, and data integrity issues.

Read the full Anomaly Insights guide →

Quick Overview¶

• Scheduled analysis (hourly, daily, weekly)
• Severity levels: Critical, Warning, Info
• Insight types: Time series anomalies, outliers, volume spikes, orphaned references
• Notifications for critical issues

Auto-Schema Discovery¶

Let your schema evolve automatically based on incoming data patterns.

Read the full Schema Discovery guide →

Quick Overview¶

• Schemaless mode: Accept any data, infer schema later
• Auto-evolving mode: Automatically detect and suggest new fields
• AI-powered type inference: Intelligently determines field types
• Review before apply: Suggestions require approval

Best Practices¶

Start Small¶

Enable AI features on non-critical structures first to understand how they work with your data.

Review AI Suggestions¶

AI suggestions are recommendations, not mandates. Always review before accepting, especially for: - Schema changes - Auto-correct fixes - Generated code

Combine Features¶

AI features work best together: - Use Schema Discovery to build your initial schema - Enable AI Validation to maintain data quality - Set up Anomaly Detection to catch issues early

Monitor Usage¶

AI features consume compute resources. Monitor your usage in the workspace dashboard and adjust feature settings if needed.

Related Documentation¶

• Structures & Records - Core data concepts
• Compute Functions - Serverless functions
• Search - Full-text search capabilities