Create memory

Create a new memory with content analysis, embedding generation, and graph storage.

Method

client.memories.create(input: CreateMemoryInput): Promise<Memory>

Parameters

interface CreateMemoryInput {
  content: string;
  layerId: string;
  metadata?: Record<string, unknown>;
}

content (string, required): The text content of the memory
layerId (string, required): Layer identifier for organizing memories
metadata (object, optional): Additional metadata (can be empty object)

Basic Example

const memory = await client.memories.create({
  layerId: "REPLACE_THIS_LAYER_ID",
  content: 'I prefer dark mode in my applications',
});

console.log('Memory created:', memory.id);

With Metadata

const memory = await client.memories.create({
  content: 'Meeting with John at 3 PM tomorrow',
  layerId: "REPLACE_THIS_LAYER_ID",
  metadata: {
    type: 'appointment',
    priority: 'high',
    participants: ['John']
  }
});

console.log('Memory created with metadata:', memory.metadata);

Response

Returns a complete Memory object with automatic analysis:

{
  id: "REPLACE_THIS_MEMORY_ID",
  layerId: "REPLACE_THIS_LAYER_ID",
  content: 'I prefer dark mode in my applications',
  embedding: [0.123, -0.456, ...], // 1536-dimensional vector
  metadata: {
    // Your custom metadata
    source: 'preferences',
    
    // Automatically added analysis
    analysis: {
      sentiment: 'neutral',
      intent: 'preference',
      topics: ['ui', 'preferences'],
      summary: 'User interface preference for dark mode',
      entities_count: 2,
      relationships_count: 1
    },
    processing: {
      analyzed_at: '2024-01-15T10:30:00Z',
      embedding_model: 'text-embedding-3-large',
      embedding_dimensions: 1536
    }
  },
  createdAt: '2024-01-15T10:30:00Z',
  updatedAt: '2024-01-15T10:30:00Z'
}

Processing Details

When you create a memory, the following processing occurs automatically:

Content Analysis: Extracts sentiment, intent, topics, entities, and relationships
Embedding Generation: Creates a 1536-dimensional vector using text-embedding-3-large
Graph Storage: Stores entities and relationships in the graph database
Enhanced Metadata: Adds analysis results and processing information

Best Practices

Layer Organization

Use layerId to organize memories by user, session, or context:

// User-specific memories
await client.memories.create({
  content: 'User prefers dark mode',
  layerId: "REPLACE_THIS_LAYER_ID"
});

// Session-specific memories
await client.memories.create({
  content: 'Working on authentication feature',
  layerId: 'REPLACE_THIS_LAYER_ID'
});

Rich Metadata

Add structured metadata for better organization:

await client.memories.create({
  content: 'Bug found in payment processing',
  layerId: "REPLACE_THIS_LAYER_ID",
  metadata: {
    type: 'issue',
    severity: 'high',
    component: 'payments',
    reportedBy: 'USER_ID',
    timestamp: new Date().toISOString()
  }
});

Error Handling

try {
  const memory = await client.memories.create({
    content: 'My important memory',
    layerId: "REPLACE_THIS_LAYER_ID",
    metadata: {}
  });
  
  console.log('Memory created:', memory.id);
} catch (error) {
  if (error.message.includes('layerId')) {
    console.error('Layer ID is required');
  } else if (error.message.includes('content')) {
    console.error('Content is required and cannot be empty');
  } else {
    console.error('Failed to create memory:', error.message);
  }
}

Create memory

On this page