Memory

Give an agent a conversation buffer that persists across turns.

Memory allows an agent to remember previous conversations.

Instead of treating every request as a new conversation, the agent can retain context and generate more natural responses.

Create a Memory#

Python

from abzagent import Memory

memory = Memory()

Use Memory with an Agent#

agent.py

from abzagent import Agent, Memory

memory = Memory()

agent = Agent(
    name="Assistant",
    instructions="You are a helpful AI assistant.",
    model="gemini-2.5-flash",
    memory=memory
)

Your agent will now remember previous messages during the conversation.

How It Works#

Every call to run() appends the new message and the agent's reply to the same Memory instance.
On the next run(), the full conversation so far is replayed to the model as context — that's how it "remembers" earlier turns.
Memory is just a buffer attached to one agent instance — nothing is written to disk unless you do it yourself.

Example#

Python

agent.run("My name is Abu Bakar.")

result = agent.run("What is my name?")

print(result.content)

Example Output

Output

Your name is Abu Bakar.

Clearing Memory#

Create a new Memory instance to start a fresh conversation.

Python

memory = Memory()

When to Use Memory#

Memory is useful for:

AI Chatbots
Customer Support
Personal Assistants
Research Assistants
Voice Assistants

Best Practices#

Create one memory instance for each conversation.
Reuse the same memory object during the conversation.
Create a new memory instance for a new conversation.

Next Step#

Continue to Tools.