Overview

Recall gives your AI applications persistent memory. Extract facts from conversations, store them in your database, and retrieve relevant context automatically.

import { createMemory } from '@youcraft/recall'

const memory = createMemory({ db, embeddings, extractor })

// Extract facts from conversations
await memory.extract(conversation, { userId: 'user_123' })

// Query relevant memories
const memories = await memory.query('user preferences', { userId: 'user_123' })

How It Works

Conversation                    Extracted Facts
─────────────                   ─────────────────
"My name is Sarah"          →   "User's name is Sarah"
"I work at Acme Corp"       →   "User works at Acme Corp"
"I love TypeScript"         →   "User loves TypeScript"

When you need context, query with natural language:

const memories = await memory.query('What programming languages?', { userId })
// → ["User loves TypeScript"]

Consolidation is the magic. When you extract “User’s name is John Doe” but already have “User’s name is John”, Recall updates the existing memory instead of creating a duplicate.

Next Steps

• Quickstart — Build a memory-enabled chatbot in 5 minutes
• Core Concepts — Learn how extraction and consolidation work
• Packages — All packages and what they do