agenticx-memory-architect

AgenticX Memory Architect

Guide for building agents with persistent memory capabilities.

Overview

AgenticX integrates with Mem0 for long-term memory, providing agents with the ability to remember past interactions, learn from experience, and maintain context across sessions.

Installation

pip install "agenticx[memory]"
# Includes: mem0, chromadb, qdrant-client, redis, milvus

Memory System Components

Component	Purpose
MemoryManager	Core memory management interface
Mem0Integration	Bridge to Mem0's memory engine
ContextMemory	Short-term, session-scoped memory
LongTermMemory	Persistent, cross-session memory

Basic Memory Usage

Initialize Memory

from agenticx.memory import MemoryManager

memory = MemoryManager(
    provider="mem0",
    config={
        "llm": {"provider": "openai", "config": {"model": "gpt-4"}},
        "vector_store": {"provider": "chroma"}
    }
)

Store and Retrieve

# Add a memory
memory.add(
    content="User prefers concise reports with bullet points",
    user_id="user-123",
    agent_id="analyst"
)

# Search memories
results = memory.search(
    query="What format does the user prefer?",
    user_id="user-123"
)
for r in results:
    print(f"[{r.score:.2f}] {r.content}")

# Get all memories for a user
all_memories = memory.get_all(user_id="user-123")

Memory-Enhanced Agents

Attach Memory to an Agent

from agenticx import Agent, AgentExecutor
from agenticx.memory import MemoryManager
from agenticx.llms import OpenAIProvider

memory = MemoryManager(provider="mem0")
agent = Agent(
    id="assistant",
    name="Personal Assistant",
    role="Assistant with memory",
    goal="Help users while remembering their preferences",
    organization_id="default"
)

executor = AgentExecutor(
    agent=agent,
    llm=OpenAIProvider(model="gpt-4"),
    memory=memory
)

# First interaction — learns preference
result = executor.run(task_1)

# Later interaction — recalls preference
result = executor.run(task_2)  # agent remembers context from task_1