Gerenciando o estado e as sessões dos agentes
Implemente métodos eficazes para manter o estado do agente entre várias interações e sessões de usuários em ambientes de produção.
Gerenciando o estado e as sessões dos agentes é uma aula grátis de AI Agents with LangChain & Autonomous Workflows no CoddyKit. Esta é a aula 2 de 4. Você pode ler a aula completa abaixo gratuitamente — depois pratica ao vivo no navegador com um editor de código integrado e um tutor de IA 24/7. Faz parte do caminho de aprendizado de AI Agents with LangChain & Autonomous Workflows, e seu progresso é sincronizado entre a web e o app CoddyKit. O curso de AI Agents with LangChain & Autonomous Workflows inclui 4 aulas no total.
Partes desta aula ainda não foram traduzidas e aparecem em inglês.
Why Agent State Matters
When building AI agents for real users, especially in production, your agent needs to remember things. Imagine a chatbot that forgets everything you said after each message – it would be frustrating!
This is where agent state and session management come in. They allow your agent to maintain context and have meaningful, continuous conversations.
Defining Agent State
Agent 'state' refers to all the information an agent needs to remember about a specific interaction or user session. This can include:
- Past messages in a conversation
- User preferences or settings
- Intermediate results from tool usage
- Any data collected during an interaction
Essentially, it's the agent's short-term and long-term memory for a given user.
The Stateless Challenge
By default, many interactions with Large Language Models (LLMs) are stateless. This means each API call is independent; the LLM doesn't inherently remember previous queries or responses.
In a production environment with many users, if you don't manage state, every interaction starts fresh. This leads to repetitive questions and a poor user experience.
Identifying User Sessions
To manage state for multiple users concurrently, we assign a unique 'session ID' to each user's interaction. This ID acts as a key to retrieve and store their specific conversation history and data.
Here's how you might generate a simple session ID:
import uuid
def start_user_session():
# Generate a unique session ID
session_id = str(uuid.uuid4())
print(f"New session started with ID: {session_id}")
return session_id
if __name__ == "__main__":
# In a real app, this ID would be tied to a user
# and sent with each request.
current_session_id = start_user_session()
# Use current_session_id to store and retrieve state
Persistent Storage for State
For production applications, in-memory storage for state is insufficient. If your server restarts, all in-memory state is lost. You need persistent storage.
Common choices include:
- Key-value stores: Redis, Memcached
- Databases: PostgreSQL, MongoDB, DynamoDB
- Cloud storage: S3 (for larger, less frequent state)
These solutions ensure state survives restarts and can be accessed across distributed services.
LangChain's Memory Abstraction
LangChain simplifies state management with its Memory modules. These modules abstract away the complexity of storing and retrieving conversation history.
The simplest is ConversationBufferMemory, which stores messages in-memory. Let's see how it keeps track of messages:
from langchain.memory import ConversationBufferMemory
from langchain_core.messages import HumanMessage, AIMessage
# Initialize in-memory conversation buffer
memory = ConversationBufferMemory(
memory_key="chat_history",
return_messages=True
)
# Simulate adding messages to the memory
memory.chat_memory.add_user_message("Hello, who are you?")
memory.chat_memory.add_ai_message("I am an AI assistant.")
memory.chat_memory.add_user_message("What can you do?")
memory.chat_memory.add_ai_message("I can answer questions and help with tasks.")
# Retrieve the current conversation history
history = memory.load_memory_variables({})["chat_history"]
print("Current conversation history:")
for message in history:
print(f"{message.type.capitalize()}: {message.content}")
Integrating External Memory
While ConversationBufferMemory is great for development, production requires external persistence. LangChain provides specialized memory classes to integrate with various backends, like Redis or databases.
You connect these external stores by passing a ChatMessageHistory object to the ConversationBufferMemory (or other memory types).
from langchain.memory import ConversationBufferMemory
from langchain_community.chat_message_histories import RedisChatMessageHistory
import os
# In a real application, you would configure Redis URL
# os.environ["REDIS_URL"] = "redis://localhost:6379/0"
# For this runnable example, we'll simulate the external history store
class SimulatedChatMessageHistory:
def __init__(self, session_id):
self.session_id = session_id
self._messages = []
print(f"\nSimulated history for session: {session_id}")
def add_user_message(self, message):
# In a real app, this would save to Redis/DB
self._messages.append(message)
print(f"[Simulated Save] User: {message}")
def add_ai_message(self, message):
# In a real app, this would save to Redis/DB
self._messages.append(message)
print(f"[Simulated Save] AI: {message}")
@property
def messages(self):
# In a real app, this would load from Redis/DB
return self._messages
# Create a unique session ID for a user
user_session_id = "user_prod_session_123"
# Initialize a simulated external history store for this session
simulated_history = SimulatedChatMessageHistory(session_id=user_session_id)
# Now, integrate this with LangChain's memory system
memory = ConversationBufferMemory(
chat_memory=simulated_history,
memory_key="history",
return_messages=True
)
# Simulate adding messages through the LangChain memory
# LangChain handles calling add_user_message/add_ai_message on simulated_history
memory.save_context({"input": "Hi there!"}, {"output": "Hello! How can I assist you?"})
memory.save_context({"input": "Tell me about AI agents."}, {"output": "AI agents combine LLMs with tools to perform tasks."})
print("\n--- Messages retrieved from LangChain memory (via simulated external store) ---")
for msg in memory.chat_memory.messages:
print(f"-> {msg.type.capitalize()}: {msg.content}")
Production Session Strategies
In production, your web framework (e.g., Flask, FastAPI, Node.js Express) will typically manage assigning and tracking session IDs for users. When a user interacts with your agent:
- The framework identifies the user's session ID.
- This ID is passed to your agent service.
- Your agent service uses the ID to load the correct conversation state from persistent storage.
- After the agent processes the request, the updated state is saved back to persistent storage using the same ID.
State Management Best Practices
To ensure robust and scalable state management in production:
- Session Expiry: Implement mechanisms to automatically clear old or inactive sessions to save storage costs and protect privacy.
- Concurrency: Design your system to handle multiple requests from the same user safely, preventing race conditions when updating state.
- Security: Protect session IDs (e.g., use secure cookies) and encrypt sensitive data stored in your persistent memory.
- Scalability: Choose a persistent store that can scale horizontally with your user base and offers low-latency access.
Check Your Understanding
Which of the following is the primary reason for using a persistent storage solution (like Redis or a database) for agent state in a production environment, rather than just in-memory storage?
Recap: State & Sessions
We've learned that managing agent state and user sessions is vital for building robust, conversational AI agents in production. Key takeaways include:
- Stateless Nature: LLMs are stateless by default, requiring explicit state management.
- Session IDs: Used to uniquely identify and manage individual user contexts.
- Persistent Storage: Essential for saving state across server restarts and distributed systems (e.g., Redis, databases).
- LangChain Memory: Provides powerful abstractions to integrate various memory backends with your agents.
Mastering state management is key to delivering seamless and intelligent agent experiences. Next, we'll explore scaling agent architectures!
Aprenda AI Agents with LangChain & Autonomous Workflows com um tutor de IA — grátis
Escreva e execute código real no seu navegador, obtenha ajuda instantânea de um tutor de IA 24/7 e continue de onde parou na web ou no app.
- Cursos
- 12
- Aulas
- 50
Perguntas Frequentes
A aula “Gerenciando o estado e as sessões dos agentes” é grátis?
Sim — o texto completo de “Gerenciando o estado e as sessões dos agentes” é grátis para ler aqui na web. Para praticá-la interativamente (um editor de código integrado e um tutor de IA 24/7) e desbloquear o restante do curso de AI Agents with LangChain & Autonomous Workflows, atualize para CoddyKit PRO. O curso de AI Agents with LangChain & Autonomous Workflows inclui 4 aulas no total.
O que vou aprender em “Gerenciando o estado e as sessões dos agentes”?
Implemente métodos eficazes para manter o estado do agente entre várias interações e sessões de usuários em ambientes de produção. Você pratica AI Agents with LangChain & Autonomous Workflows com código prático que executa diretamente no navegador, e um tutor de IA 24/7 responde suas dúvidas enquanto trabalha na aula.
Preciso ter experiência prévia para começar AI Agents with LangChain & Autonomous Workflows?
Nenhuma experiência prévia é necessária. AI Agents with LangChain & Autonomous Workflows no CoddyKit é estruturado para alunos iniciantes até avançados, então você pode começar aqui ou desde o início e aprender no seu ritmo. Esta é a aula 2 de 4.
Quanto tempo leva a aula “Gerenciando o estado e as sessões dos agentes”?
A maioria das aulas CoddyKit leva cerca de 5–10 minutos. Cada uma é compacta e interativa, então você faz progresso constante e retoma exatamente de onde parou entre web e app.
Posso escrever e executar código nesta aula de AI Agents with LangChain & Autonomous Workflows?
Sim. Cada aula de AI Agents with LangChain & Autonomous Workflows inclui um editor de código integrado, então você escreve e executa código real direto no navegador e recebe feedback de IA instantaneamente — nenhuma configuração local necessária.
Todas as aulas deste curso
- Implantando agentes em plataformas de nuvem
- Gerenciando o estado e as sessões dos agentes
- Dimensionando arquiteturas de agentes
- Limitação de taxa e gerenciamento de cotas da API