Agent Memory

Understanding how agents maintain context and learn

Memory lets agents remember conversations and learn from past interactions. Without memory, every message is like starting fresh.

---

Quick Start

from praisonaiagents import Agent

# Enable memory with one parameter
agent = Agent(
    instructions="You are a personal assistant",
    memory=True  # Remember conversations
)

agent.start("My name is Alex and I love Python")
# Later...
agent.start("What's my name?")  # Agent remembers: "Alex"

Just add memory=True to enable conversation memory.

---

Types of Memory

Short-Term

Remembers current chat

Long-Term

Persists across sessions

Knowledge

Access to documents

---

Memory Options

1. Basic Memory (Conversation)

   agent = Agent(
       instructions="You are helpful",
       memory=True  # Remembers the conversation
   )
   
2. Long-Term Memory (Persistent)

   agent = Agent(
       instructions="You are a personal assistant",
       memory={
           "provider": "sqlite",  # Store in database
           "db_path": "memory.db"
       }
   )
   
3. Knowledge Memory (Documents)

   agent = Agent(
       instructions="Answer from documents",
       knowledge=["docs/"],  # Load documents
       memory=True
   )
   

---

How Memory Works

Step	What Happens
Input	You send a message
Retrieve	Agent checks memory for context
Process	Combines memory + new input
Respond	Generates informed response
Save	Stores important info for later

---

Complete Example

from praisonaiagents import Agent

# Personal assistant with memory
assistant = Agent(
    name="PersonalAssistant",
    instructions="""You are a personal assistant.
    Remember user preferences and past conversations.
    Be helpful and personalized.""",
    memory=True
)

# First conversation
assistant.start("I prefer short, bullet-point answers")
assistant.start("My favorite programming language is Python")

# Later conversation - agent remembers!
assistant.start("Explain how to read a file")
# Agent will give short, bullet-point answer about Python

---

Multi-Agent Memory

Agents can share memory or have private memory:

from praisonaiagents import Agent, AgentTeam

# Agents with shared memory
researcher = Agent(instructions="Research topics", memory=True)
writer = Agent(instructions="Write summaries", memory=True)

team = AgentTeam(
    agents=[researcher, writer],
    memory=True  # Shared team memory
)

---

Best Practices

Start Simple

Use memory=True first, add complexity later

Be Specific

Tell the agent what to remember in instructions

Use Knowledge

For documents, use knowledge= not memory

Test Memory

Verify the agent remembers correctly

---

Next: Multi-Agent Systems

Learn how multiple agents work together.