Core Components

Memory System

Vector-backed memories table with arbitrary memoryType strings. Common types include agent_learning, solution_pattern, and codebase_knowledge, but any string is valid. Vector search for semantic retrieval.

Common Memory Types

The memoryType field is an arbitrary string. These are the commonly used types:

memoryTypePurpose
agent_learningWhat the agent learned from a task
solution_patternReusable patterns discovered
codebase_knowledgeFacts about the codebase

Any string is valid as a memoryType -- these are not a fixed enum.

API Usage

import { createLibSQLLocalDataLayer } from '@jetpack-agent/data';

const dataLayer = await createLibSQLLocalDataLayer('.jetpack/jetpack.db');

// Store a memory
await dataLayer.memories.store({
  agentId: 'agent-1',
  taskId: 'task-1',
  content: 'Implemented auth using JWT with 24h expiry',
  memoryType: 'agent_learning',  // any string, e.g. 'agent_learning', 'solution_pattern', 'codebase_knowledge'
  importance: 0.8,  // 0-1 scale
  tags: ['backend', 'security'],
});

// Get memories relevant to a task
const memories = await dataLayer.memories.getTaskContext(taskId);

// Store a solution pattern
await dataLayer.memories.store({
  agentId: 'agent-1',
  taskId: 'task-2',
  content: 'All API endpoints follow RESTful conventions with /api/v1 prefix',
  memoryType: 'solution_pattern',
  importance: 0.8,
  tags: ['api-design'],
});

// Store codebase knowledge
await dataLayer.memories.store({
  agentId: 'agent-1',
  taskId: 'task-3',
  content: 'The auth module uses bcrypt for password hashing',
  memoryType: 'codebase_knowledge',
  importance: 0.9,
  tags: ['auth', 'security'],
});

Configuration

// Memory types available:
// 'agent_learning'      — What the agent learned from a task
// 'solution_pattern'    — Reusable patterns discovered
// 'codebase_knowledge'  — Facts about the codebase

// Field name is 'memoryType', not 'type'
// Tags are a string array, not nested metadata