LangChain QA BotでAWSリソースの使い所を考えてみる

!pip install boto3 openai langchain llama-index

!git clone https://github.com/jerryjliu/gpt_index.git
%cd gpt_index/examples/paul_graham_essay/data
!ls     # paul_graham_essay.txt

import os
os.environ["OPENAI_API_KEY"] = "xxxxxxxxxxxx"

from llama_index import GPTSimpleVectorIndex, SimpleDirectoryReader

documents = SimpleDirectoryReader('data').load_data()
index = GPTSimpleVectorIndex.from_documents(documents)
index.save_to_disk('sample_index.json')

from pathlib import Path
import boto3
import io

s3 = boto3.resource('s3')
bucket = s3.Bucket('sample_bucket')

KEY = 'sample_index.json'
ORIGIN_PATH = Path('sample_index.json')

bucket.upload_file(str(ORIGIN_PATH), KEY)

import boto3
from langchain.agents import initialize_agent, Tool
from langchain.tools import BaseTool
from langchain.llms import OpenAI
from llama_index import GPTSimpleVectorIndex

s3 = boto3.resource('s3')
json = s3.Bucket("sample_bucket").Object("sample_index.json").get()['Body'].read()

index = GPTSimpleVectorIndex.load_from_string(json)

tools = [
    Tool(
        name="Paul Graham Essay",
        func=lambda q: str(index.query(q)),
        description="Useful for the generating the answers about Paul Graham",
        return_direct=False
    ),
]

llm = OpenAI(temperature=0)
agent = initialize_agent(tools, llm, agent="zero-shot-react-description", verbose=True)

agent.run("What did the author do growing up?")

from langchain.memory import ConversationBufferMemory
from langchain.memory.chat_message_histories import DynamoDBChatMessageHistory

message_history = DynamoDBChatMessageHistory(table_name="langchain-sample", session_id='s000001')
message_history.add_user_message("hi!")
message_history.add_ai_message("whats up?")
message_history.add_user_message("I'm great. How about you?")
message_history.add_ai_message("I'm great too. Thanks!")

message_history.messages

[HumanMessage(content='hi!', additional_kwargs={}),
 AIMessage(content='whats up?', additional_kwargs={}),
 HumanMessage(content="I'm great. How about you?", additional_kwargs={}),
 AIMessage(content="I'm great too. Thanks!", additional_kwargs={})]

import boto3

dynamodb = boto3.resource('dynamodb')
table = dynamodb.Table('langchain-sample')

ret = table.get_item(Key={"SessionId": "s000001"})
ret["Item"]

{'SessionId': 's000001',
 'History': [{'type': 'human',
   'data': {'content': 'hi!', 'additional_kwargs': {}}},
  {'type': 'ai', 'data': {'content': 'whats up?', 'additional_kwargs': {}}},
  {'type': 'human',
   'data': {'content': "I'm great. How about you?", 'additional_kwargs': {}}},
  {'type': 'ai',
   'data': {'content': "I'm great too. Thanks!", 'additional_kwargs': {}}}]}

from langchain.llms import OpenAI
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferMemory
from langchain.memory.chat_message_histories import DynamoDBChatMessageHistory
import os

os.environ["OPENAI_API_KEY"] = "xxxxxxxxxxxx"

llm = OpenAI(temperature=0)

message_history = DynamoDBChatMessageHistory(table_name="langchain-sample", session_id='s000001')

memory = ConversationBufferMemory(chat_memory=message_history)

conversation = ConversationChain(
    llm=llm, 
    verbose=True, 
    memory=memory
)

conversation.predict(input="Hi there!")

> Entering new ConversationChain chain...
Prompt after formatting:
The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.

Current conversation:

Human: Hi there!
AI:

> Finished chain.
" Hi there! It's nice to meet you. How can I help you today?"

conversation.predict(input="How's weather there?")

> Entering new ConversationChain chain...
Prompt after formatting:
The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.

Current conversation:
Human: Hi there!
AI:  Hi there! It's nice to meet you. How can I help you today?
Human: How's weather there?
AI:

> Finished chain.
" The weather here is sunny and warm. The temperature is currently 75 degrees Fahrenheit and the humidity is low. It's a great day for outdoor activities!"

{'SessionId': 's000001',
 'History': [{'type': 'human',
   'data': {'content': 'Hi there!', 'additional_kwargs': {}}},
  {'type': 'ai',
   'data': {'content': " Hi there! It's nice to meet you. How can I help you today?",
    'additional_kwargs': {}}},
  {'type': 'human',
   'data': {'content': "How's weather there?", 'additional_kwargs': {}}},
  {'type': 'ai',
   'data': {'content': " The weather here is sunny and warm. The temperature is currently 75 degrees Fahrenheit and the humidity is low. It's a great day for outdoor activities!",
    'additional_kwargs': {}}}]}

from langchain.document_loaders import UnstructuredFileLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import FAISS

loader = UnstructuredFileLoader("sample.txt")
documents = loader.load()
text_splitter = RecursiveCharacterTextSplitter(chunk_size = 500, chunk_overlap = 100, length_function = len,)
docs = text_splitter.split_documents(documents)
embeddings = OpenAIEmbeddings()
db = FAISS.from_documents(docs, embeddings)
db.save_local("sample_index")

!ls -lt 

drwxrwxr-x 2 ec2-user ec2-user   4096 Apr 30 14:06 sample_index

! ls -lt sample_index

-rw-rw-r-- 1 ec2-user ec2-user 129069 Apr 30 14:09 index.faiss
-rw-rw-r-- 1 ec2-user ec2-user  30726 Apr 30 14:09 index.pkl

import os
import boto3
from langchain.document_loaders import UnstructuredFileLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import FAISS

bucket_name = "sample_bucket"
orig_docs_dir = "orig_docs"
file_name = "sample.txt"

os.environ['OPENAI_API_KEY'] = "xxxxxxxxxx"

def upload_dir_s3(dirpath, s3bucket):
    for root,dirs,files in os.walk(dirpath):
        for file in files:
            s3bucket.upload_file(os.path.join(root,file),os.path.join(root,file))

def download_dir_s3(dirpath, s3bucket):
    for obj in s3bucket.objects.filter(Prefix = dirpath):
        if not os.path.exists(os.path.dirname(obj.key)):
            os.makedirs(os.path.dirname(obj.key))
        s3bucket.download_file(obj.key, obj.key)
        
s3 = boto3.resource('s3')
s3_bucket = s3.Bucket(bucket_name)
download_dir_s3(orig_docs_dir, s3_bucket)

loader = UnstructuredFileLoader(f"{orig_docs_dir}/{file_name}")
documents = loader.load()
text_splitter = RecursiveCharacterTextSplitter(chunk_size = 500, chunk_overlap = 100, length_function = len,)
docs = text_splitter.split_documents(documents)
embeddings = OpenAIEmbeddings()
db = FAISS.from_documents(docs, embeddings)
db.save_local("sample_index")

upload_dir_s3("sample_index", s3_bucket)

import os
import boto3
from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import FAISS

bucket_name = "my-sagemaker-notebook-bucket"
index_dir = "sample_index"

os.environ['OPENAI_API_KEY'] = "xxxxxxxxx"

def download_dir_s3(dirpath, s3bucket):
    for obj in s3bucket.objects.filter(Prefix = dirpath):
        if not os.path.exists(os.path.dirname(obj.key)):
            os.makedirs(os.path.dirname(obj.key))
        s3bucket.download_file(obj.key, obj.key)
        
s3 = boto3.resource('s3')
s3_bucket = s3.Bucket(bucket_name)
download_dir_s3(index_dir, s3_bucket)

embeddings = OpenAIEmbeddings()
db = FAISS.load_local(index_dir, embeddings)

(・・・)