`createShortTermMemory`

Problem

In LangChain agents, conversation history is usually a plain array you manage manually — pushing messages, slicing old ones, passing it around between functions. There is no standard interface, so every project does it differently and bugs creep in.

Solution

createShortTermMemory creates a typed in-memory message store for the duration of a session. It gives you a clean API to add, read, and clear messages — so your conversation management is consistent and explicit across your Node.js application.

Feature & Use-Case

Use createShortTermMemory when:

You need per-session conversation history (one object per user, per request, per thread)
You want a clean API instead of pushing to raw arrays
You are building a chat agent or multi-turn assistant where messages accumulate in-memory
You do not need persistence across restarts — for that, use createLongTermMemory

Import


import { createShortTermMemory } from "llm-layer-engine";

Function Signature


function createShortTermMemory(): ShortTermMemory

`ShortTermMemory` Interface


interface ShortTermMemory {
  add(message: Message): void;
  addMany(messages: Message[]): void;
  getAll(): Message[];
  clear(): void;
  size(): number;
}

Method Reference

Method	Description
`add(message)`	Add a single message to memory
`addMany(messages)`	Add multiple messages at once
`getAll()`	Return a copy of all stored messages
`clear()`	Remove all messages from memory
`size()`	Return the current message count

Example — Chat Session with Memory


import { createShortTermMemory, createAgent } from "llm-layer-engine";
import type { LLMProvider } from "llm-layer-engine";
 
const memory = createShortTermMemory();
const agent = createAgent({ provider: myProvider, model: "claude-3-5-sonnet-20241022" });
 
// Turn 1 — User sends first message
memory.add({ role: "user", content: "My name is Burhan." });
 
let { result } = await agent.run(memory.getAll());
memory.add({ role: "assistant", content: result });
 
// Turn 2 — User asks a follow-up
memory.add({ role: "user", content: "What is my name?" });
 
({ result } = await agent.run(memory.getAll()));
memory.add({ role: "assistant", content: result });
 
console.log(result);       // → "Your name is Burhan."
console.log(memory.size()); // → 4

Example — Reset Memory Between Sessions


const memory = createShortTermMemory();
 
// Session A
memory.add({ role: "user", content: "Hello from session A" });
console.log(memory.size()); // 1
 
// End session — clear before next user
memory.clear();
console.log(memory.size()); // 0

Conclusion

createShortTermMemory keeps your conversation history clean and typed. Create one instance per session, feed memory.getAll() into every agent call, and call memory.clear() when the session ends. It is the simplest way to build a stateful chat agent in your Node.js + LangChain project.

createShortTermMemory