Code can be found here: GitHub - jamesbmour/blog_tutorials:
In the ever-evolving world of conversational AI and language models, maintaining context and efficiently managing information flow are critical components of building intelligent applications. LangChain, a powerful framework designed for working with large language models (LLMs), offers robust tools for memory management and data persistence, enabling the creation of context-aware systems.
In this guide, we'll delve into the nuances of leveraging memory and storage in LangChain to build smarter, more responsive applications.
Memory management in LangChain allows applications to retain context, making interactions more coherent and contextually relevant. Let’s explore the different memory types and their use cases.
LangChain provides various memory types to address different scenarios. Here, we’ll focus on two key types:
ConversationBufferMemory
This memory type is ideal for short-term context retention, capturing and recalling recent interactions in a conversation.
from langchain.memory import ConversationBufferMemory
memory = ConversationBufferMemory()
memory.save_context({"input": "Hi, I'm Alice"}, {"output": "Hello Alice, how can I help you today?"})
memory.save_context({"input": "What's the weather like?"}, {"output": "I'm sorry, I don't have real-time weather information. Is there anything else I can help you with?"})
print(memory.load_memory_variables({}))
ConversationSummaryMemory
For longer conversations, ConversationSummaryMemory is a great choice. It summarizes key points, maintaining context without overwhelming detail.
from langchain.memory import ConversationSummaryMemory
from langchain.llms import Ollama
llm = Ollama(model='phi3',temperature=0)
memory = ConversationSummaryMemory(llm=llm)
memory.save_context({"input": "Hi, I'm Alice"}, {"output": "Hello Alice, how can I help you today?"})
memory.save_context({"input": "I'm looking for a good Italian restaurant"}, {"output": "Great! I'd be happy to help you find a good Italian restaurant. Do you have any specific preferences or requirements, such as location, price range, or specific dishes you're interested in?"})
print(memory.load_memory_variables({}))
Selecting the appropriate memory type depends on several factors:
Use Cases:
Memory can be seamlessly integrated into LangChain chains and agents to enhance conversational capabilities.
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferMemory
# llm = OpenAI(temperature=0)
memory = ConversationBufferMemory()
conversation = ConversationChain(
llm=llm,
memory=memory,
verbose=True
)
conversation.predict(input="Hi, I'm Alice")
conversation.predict(input="What's my name?")
This example illustrates how ConversationBufferMemory can be used to remember previous interactions, enabling more natural conversations.
Persistent storage ensures that conversation history and context are maintained across sessions, enabling continuity in interactions.
For basic persistence, you can use file-based storage with JSON:
import json
class PersistentMemory:
def __init__(self, file_path):
self.file_path = file_path
self.load_memory()
def load_memory(self):
try:
with open(self.file_path, 'r') as f:
self.chat_memory = json.load(f)
except FileNotFoundError:
self.chat_memory = {'messages': []}
def save_memory(self):
with open(self.file_path, 'w') as f:
json.dump({'messages': self.chat_memory['messages']}, f)
# Usage
memory = PersistentMemory(file_path='conversation_history.json')
print(memory.chat_memory)
This method allows you to persist conversation history in a simple, human-readable format.
For more scalable and efficient storage, integrating with databases like SQLite is recommended:
import sqlite3
class SQLiteMemory:
def __init__(self, db_path):
self.db_path = db_path
self.conn = sqlite3.connect(db_path)
self.create_table()
def create_table(self):
cursor = self.conn.cursor()
cursor.execute('''
CREATE TABLE IF NOT EXISTS conversations
(id INTEGER PRIMARY KEY, input TEXT, output TEXT)
''')
self.conn.commit()
def save_context(self, inputs, outputs):
cursor = self.conn.cursor()
cursor.execute('INSERT INTO conversations (input, output) VALUES (?, ?)',
(inputs['input'], outputs['output']))
self.conn.commit()
def load_memory_variables(self, inputs):
cursor = self.conn.cursor()
cursor.execute('SELECT input, output FROM conversations ORDER BY id DESC LIMIT 10')
rows = cursor.fetchall()
history = "\\n".join([f"Human: {row[0]}\\nAI: {row[1]}" for row in reversed(rows)])
return {"history": history }
# Usage
memory = SQLiteMemory('conversation_history.db')
print(memory.load_memory_variables({}))
To ensure your application remains responsive, consider these optimization strategies:
Here’s an example of a memory class with basic caching:
import time
class CachedSQLiteMemory(SQLiteMemory):
def __init__(self, db_path, cache_ttl=60):
super().__init__(db_path)
self.cache = None
self.cache_time = 0
self.cache_ttl = cache_ttl
def load_memory_variables(self, inputs):
current_time = time.time()
if self.cache is None or (current_time - self.cache_time) > self.cache_ttl:
var = self.cache
self.cache = super().load_memory_variables(inputs)
self.cache_time = current_time
return self.cache
memory = CachedSQLiteMemory('conversation_history.db', cache_ttl=30)
This implementation caches the results of database queries for a specified time, reducing the load on the database and improving performance for applications that frequently access memory data.
Effective memory management is a cornerstone of building intelligent, context-aware conversational AI applications. LangChain provides a flexible and powerful framework for managing memory, allowing developers to tailor memory types to specific use cases, implement persistent storage solutions, and optimize performance for large-scale applications.
透過選擇正確的記憶體類型、整合持久性儲存並利用自訂記憶體類別和快取策略等先進技術，您可以建立複雜的AI 系統，即使在規模和複雜性方面也能維護上下文、改善使用者體驗並高效運作互動次數不斷增加。
有了這些可用的工具和技術，您就可以充分利用 LangChain 的全部潛力來創建響應式、智慧和情境感知的人工智慧應用程式。無論您是開發客戶支援機器人、虛擬助理，還是複雜的對話系統，掌握 LangChain 中的記憶體和儲存都將是您成功的關鍵因素。
如果你想支持我的寫作或買瓶啤酒給我：
https://buymeacoffee.com/bmours
以上是LangChain部分利用LangChain中的記憶體和儲存：綜合指南的詳細內容。更多資訊請關注PHP中文網其他相關文章！
