What is short term memory in AI agents
AI agents is a temporary storage of recent interactions or context that the agent uses to maintain coherent conversations and make decisions. It differs from long term memory by focusing only on immediate, short-lived information relevant to the current task or dialogue.How it works
Short term memory in AI agents acts like a scratchpad that holds recent inputs, outputs, and relevant context during an ongoing interaction. Imagine a detective taking quick notes during an interview to remember key facts before writing a full report later. This memory is volatile and resets or updates frequently as the conversation or task progresses, enabling the agent to respond consistently without needing to reprocess all prior data.
Concrete example
Here is a simple Python example simulating short term memory in an AI chat agent using a list to store recent messages:
import os
from openai import OpenAI
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
class ShortTermMemoryAgent:
def __init__(self, max_memory=3):
self.memory = []
self.max_memory = max_memory
def remember(self, message):
self.memory.append(message)
if len(self.memory) > self.max_memory:
self.memory.pop(0) # Remove oldest memory
def chat(self, user_input):
self.remember(f"User: {user_input}")
prompt = "\n".join(self.memory) + "\nAgent:"
response = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": prompt}]
)
answer = response.choices[0].message.content
self.remember(f"Agent: {answer}")
return answer
agent = ShortTermMemoryAgent()
print(agent.chat("Hello, who won the world series in 2020?"))
print(agent.chat("And who was the MVP?"))User: Hello, who won the world series in 2020? Agent: The Los Angeles Dodgers won the World Series in 2020. User: And who was the MVP? Agent: Corey Seager was the MVP of the 2020 World Series.
When to use it
Use short term memory in AI agents when you need to maintain context over a short conversation or task session, such as chatbots, virtual assistants, or interactive agents. It is not suitable for storing persistent knowledge or facts that must survive across sessions; that requires long term memory or external databases. Short term memory improves coherence and relevance in dynamic interactions.
Key terms
| Term | Definition |
|---|---|
| Short term memory | Temporary storage of recent context or interactions in an AI agent. |
| Long term memory | Persistent storage of knowledge or data across sessions or tasks. |
| AI agent | An autonomous system that perceives and acts in an environment to achieve goals. |
| Context window | The amount of recent text or data an LLM can consider at once. |
Key Takeaways
- Short term memory holds recent context to keep AI agent responses coherent and relevant.
- It is volatile and limited in size, unlike long term memory which stores persistent knowledge.
- Implement short term memory as a rolling buffer of recent interactions for chatbots and assistants.