# 1/16 LangChain for LLM Application Development
[TOC]
---
Write by 陳正哲
## 課程大綱
主要學習內容
* Langchain屬於Knowledge retrival類的應用，不同於直接調用LLM內的knowledge，而是調用LLM模型並提供提示及資料庫解析來生成回應。
* LLM記憶體功能：存儲對話並管理有限的上下文空間。
* 鏈接功能：創建操作序列。
* 針對文件問答：應用LLM於特定數據或資料集。
* 代理人(agents)：探索作為推理代理的LLM的新開發
### Introduction
參考基本教學連結
https://pub.aimind.so/langchain-all-you-need-to-know-beadc2c93323
DeepLearning.AI課程[連結](https://learn.deeplearning.ai/langchain/lesson/1/introduction)
:::info
2022年早期ChatGPT僅提供簡單問畫面且對話token限制大(4000 tokens)，當時若要進一步客製化，則需要透過OpenAI API來開發。
LangChain 是開放原始碼架構，使用者可選擇大型語言模型(LLM)與模組打造應用程式。利用LangChain框架擴展語言模型在應用開發中的用途與能力。
LangChain 提供標準化模組例如Memory、Chains、Document loader、agents等，藉以改善模型生成資訊的客製化程度、準確度和關聯性。
:::
#### 練習小技巧(1): Colab起手式(update:2024/01)
```
#設定可用openai version
! pip install openai==0.28
! pip install python-dotenv #local computer environment
! pip install langchain
! pip install langchain_experimental
#新版agent部分功能移到此處
```
### L1-基本模組Models, Prompts and parsers
:::info
* 學習如何有效提供提示給LLM，掌握解析LLM回應的技巧
* LangChain[支援的LLMs](https://python.langchain.com/docs/integrations/llms/), 如: OpenAI, HuggingFaceHub等
* 應用程式呼叫LLM時, 不會直接將使用者的本文傳到LLM, 而是會將使用者撰寫的文字進行組合, 再傳送至LLM
:::
DeepLearning.AI[課程連結 L1](https://learn.deeplearning.ai/langchain/lesson/2/models,-prompts-and-parsers)
#### 使用API calls through LangChain
```
import os
import openai
from dotenv import load_dotenv, find_dotenv
_ = load_dotenv(find_dotenv()) # read local .env file
openai.api_key = os.environ['OPENAI_API_KEY']
```
```
# 按照使用時間點選擇GPT模型
# Get the current date
current_date = datetime.datetime.now().date()
# Define the date after which the model should be set to "gpt-3.5-turbo"
target_date = datetime.date(2024, 6, 12)
# Set the model variable based on the current date
if current_date > target_date:
llm_model = "gpt-3.5-turbo"
else:
llm_model = "gpt-3.5-turbo-0301"
```
其餘按照課程進行就可以了解基本對話模式
### L2-Memory
:::info
探索如何使用Langchain記憶體來存儲對話
瞭解管理有限上下文空間(tokens)的方法
:::
DeepLearning.AI[課程連結 L2](https://learn.deeplearning.ai/langchain/lesson/3/memory)
#### 2-1 ConversationBufferMemory
```
from langchain.chat_models import ChatOpenAI
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferMemory
```
```
llm = ChatOpenAI(temperature=0.0, model=llm_model)
memory = ConversationBufferMemory()
conversation = ConversationChain(
llm=llm,
memory = memory,
verbose=True
)
```
```
conversation.predict(input="你好，我是小強")
conversation.predict(input="What is 1+1?")
```
```
print(memory.buffer)
```
說明: 顯示memory內容
> Human: 你好，我是小強
AI: 你好小強，很高興認識你。我是AI。有什麼我可以幫助你的嗎？ (Translation: Hello Xiao Qiang, nice to meet you. I'm AI. How can I assist you?)
Human: What is 1+1?
AI: The answer to 1+1 is 2.
```
memory = ConversationBufferMemory()
```
說明: 清除memory
```
memory.load_memory_variables({})
```
自己執行看看
#### 2-2 ConversationBufferWindowMemory
```
from langchain.memory import ConversationBufferWindowMemory
```
```
memory = ConversationBufferWindowMemory(k=1)
```
說明: 設定只記最後一則對話(k=1)
```
#建立範例
memory.save_context({"input": "Hi"},
{"output": "What's up"})
memory.save_context({"input": "Not much, just hanging"},
{"output": "Cool"})
```
```
llm = ChatOpenAI(temperature=0.0, model=llm_model)
memory = ConversationBufferWindowMemory(k=1)
conversation = ConversationChain(
llm=llm,
memory = memory,
verbose=False
)
```
#第一則對話
conversation.predict(input="Hi, my name is Andrew")
#第二則對話
conversation.predict(input="What is 1+1?")
```
conversation.predict(input="What is my name?")
memory.load_memory_variables({})
```
自己執行試試看memory內容
#### 2-3 ConversationTokenBufferMemory
```
from langchain.memory import ConversationTokenBufferMemory
from langchain.llms import OpenAI
llm = ChatOpenAI(temperature=0.0, model=llm_model)
```
```
memory = ConversationTokenBufferMemory(llm=llm, max_token_limit=50)
#建立文本範例
memory.save_context({"input": "AI is what?!"},
{"output": "Amazing!"})
memory.save_context({"input": "Backpropagation is what?"},
{"output": "Beautiful!"})
memory.save_context({"input": "Chatbots are what?"},
{"output": "Charming!"})
```
說明: max_token_limit=50設定記憶為最新/近的50個tokens
```
memory.load_memory_variables({})
```
自己執行看看
#### 2-4 ConversationSummaryMemory
```
from langchain.memory import ConversationSummaryBufferMemory
```
```
# 創建超長文本範例
schedule = "There is a meeting at 8am with your product team. \
You will need your powerpoint presentation prepared. \
9am-12pm have time to work on your LangChain \
project which will go quickly because Langchain is such a powerful tool. \
At Noon, lunch at the italian resturant with a customer who is driving \
from over an hour away to meet you to understand the latest in AI. \
Be sure to bring your laptop to show the latest LLM demo."
memory = ConversationSummaryBufferMemory(llm=llm, max_token_limit=100)
memory.save_context({"input": "Hello"}, {"output": "What's up"})
memory.save_context({"input": "Not much, just hanging"},
{"output": "Cool"})
memory.save_context({"input": "What is on the schedule today?"},
{"output": f"{schedule}"})
#顯示memory
memory.load_memory_variables({})
```
```
conversation = ConversationChain(
llm=llm,
memory = memory,
verbose=True
)
#對話
conversation.predict(input="What would be a good demo to show?")
#看結果
memory.load_memory_variables({})
```
說明: 自己用文本試驗體會
### L3-Chains
:::info
學習創建操作Chain(鍊/序列)的技術
理解如何將不同操作工具串連以增強LLM的功能，如果是明確的工作流程，那麼可以考慮Chain 應用
:::
DeepLearning.AI[課程連結 L3](https://learn.deeplearning.ai/langchain/lesson/4/chains)
#### 3-1 LLMChain
```
from langchain.chat_models import ChatOpenAI
from langchain.prompts import ChatPromptTemplate
from langchain.chains import LLMChain
```
說明: 這裡會用到chains，其他看課程連結。可以直接在DeepLearning.AI的Jupyter網頁環境執行。如果要在Colab或自己電腦練習，要從網頁環境Jupyter下載Data.csv。
#### 3-2 Sequential Chains
把不同的prompt_template以chains串起
##### SimpleSequentialChain
(1)一次串聯數個prompts，每次單一輸入單一輸出
最後輸出一個結果
##### SequentialChain
(2)一次匯入多組prompts，用不同流程組合輸出結果
#### 3-3 Router Chain
(3) 運用LLM把對話關鍵字分類分流
```
from langchain.chains.router import MultiPromptChain
from langchain.chains.router.llm_router import LLMRouterChain,RouterOutputParser
from langchain.prompts import PromptTemplate
```
### L4-Question and Answer
:::info
透過Langchain應用LLM於特定數據資料，不再只是擷取LLM訓練的背景資料庫。
LangChain把大文本分解為chunks存到向量資料庫(VectorStore)，再以prompt訪問VectorStore進行問答。
學習如何自定義LLM以適應特定需求
:::
DeepLearning.AI[課程連結 L4](https://learn.deeplearning.ai/langchain/lesson/5/question-and-answer)
課程說明- 課程範例會用到OutdoorClothingCatalog_1000.csv，建議直接在網頁環境執行
(1)LLMs每次輸入的tokens有限
#### Embedding
(2) 把文本轉換為向量資料庫的過程
#### Vector Database
(1)
(2)
### L5-Evaluation
:::info
Langchain提供檢視生成回應的品質檢查工具
:::
DeepLearning.AI[課程連結 L5-Evaluation](https://learn.deeplearning.ai/langchain/lesson/6/evaluation)
(正哲: 這部分雖然可以在網頁環境執行，但生成的檢查是否這樣就足夠，可能要看case而定)
### L6-Agents
:::info
探討LLM作為agents(推理代理人)的概念和應用
瞭解如何開發Agents來實現複雜任務
範例: 使用Wikipedia、取得時間
:::
DeepLearning.AI[課程連結 L6](https://learn.deeplearning.ai/langchain/lesson/7/agents)
課程說明- 直接DeepLearning.AI的網頁環境執行沒問題。若在Colab環境需要修改如下：
```
#部分功能搬到langchain_experimental
! pip install langchain langchain_experimental
```
```
from langchain_experimental.agents.agent_toolkits import create_python_agent
from langchain.agents import load_tools, initialize_agent
from langchain.agents import AgentType
from langchain_experimental.tools.python.tool import PythonREPLTool
from langchain.python import PythonREPL
from langchain.chat_models import ChatOpenAI
```
## 實際應用和案例研究
## 問題討論與反思
> 建議大家到Langchain community與Langchain
參考這篇也有這堂課的完整介紹[LangChain Overview](https://pub.aimind.so/langchain-all-you-need-to-know-beadc2c93323)
## 參考-OpenAI新功能Assistants API
**OpenAI參考資料[連結](https://)**
**Step 1: 創建助手**
```
assistant = client.beta.assistants.create(
name="Math Tutor",
instructions="You are a personal math tutor. Write and run code to answer math questions.",
tools=[{"type": "code_interpreter"}],
model="gpt-4-1106-preview"
)
```
**Step 2: 創建Thread(連續對話鍊)**
```
thread = client.beta.threads.create()
```
**Step 3: 加入使用者訊息**
```
message = client.beta.threads.messages.create(
thread_id=thread.id,
role="user",
content="I need to solve the equation `3x + 11 = 14`. Can you help me?"
)
```
**Step 4: 執行**
```
run = client.beta.threads.runs.create(
thread_id=thread.id,
assistant_id=assistant.id,
instructions="Please address the user as Jane Doe. The user has a premium account."
)
```
Sign in with Wallet
Connect another wallet