memory types

Agents use different memory mechanisms: in-context (conversation history), external (databases, files), semantic (vector embeddings), and procedural (learned skills).

Syntax

ai-agents

in-context: messages[]
external: db.get(key) / db.set(key, value)
semantic: vector_store.search(query)

Example

ai-agents

// Agent with memory:
const memory = {
  // Short-term: conversation history
  shortTerm: [],
  
  // Long-term: key facts stored in DB
  longTerm: new Map(),
  
  // Semantic: searchable past interactions
  async searchSimilar(query) {
    return await vectorStore.similaritySearch(query, 3);
  }
};

// After each interaction:
memory.shortTerm.push({ role: "user", content: userMsg });