圣诞老人遇见 GenAI：利用大语言模型、LangChain 和 Elasticsearch 破译手写的圣诞信件

本文介绍: 在北极的中心地带，圣诞老人的精灵团队面临着巨大的后勤挑战：如何处理来自世界各地儿童的数百万封信件。圣诞老人表情坚定，他决定是时候将人工智能纳入圣诞节行动了。圣诞老人坐在配备了最新人工智能技术的电脑前，开始在 Jupyter Notebook 中编写 Python 脚本。我们的目标既简单又雄心勃勃：利用生成式人工智能和 LLM 的力量来解释手写字母并提取必要的数据，并将其组织在 Elasticsearch 中。

在北极的中心地带，圣诞老人的精灵团队面临着巨大的后勤挑战：如何处理来自世界各地儿童的数百万封信件。圣诞老人表情坚定，他决定是时候将人工智能纳入圣诞节行动了。

圣诞老人坐在配备了最新人工智能技术的电脑前，开始在 Jupyter Notebook 中编写 Python 脚本。我们的目标既简单又雄心勃勃：利用生成式人工智能和 LLM 的力量来解释手写字母并提取必要的数据，并将其组织在 Elasticsearch 中。

pip3 install python-dotenv elasticsearch langchain openai

$ pwd
/Users/liuxg/python/elser
$ cat .env
ES_USER="elastic"
ES_PASSWORD="yarOjyX5CLqTsKVE3v*d"
ES_ENDPOINT="localhost"
OPENAI_API_KEY="YourOwnOpenAiKey"

$ pwd
/Users/liuxg/python/elser
$ cp ~/elastic/elasticsearch-8.11.0/config/certs/http_ca.crt .
$ ls http_ca.crt 
http_ca.crt

$ pwd
/Users/liuxg/python/elser
$ jupyter notebook

import os

from dotenv import load_dotenv

load_dotenv()

# OpenAI API Key
OPENAI_API_KEY = os.getenv('OPENAI_API_KEY')
OPENAI_API_URL = "https://api.openai.com/v1/chat/completions"

elastic_user=os.getenv('ES_USER')
elastic_password=os.getenv('ES_PASSWORD')
elastic_endpoint=os.getenv("ES_ENDPOINT")
openai_api_key=os.getenv('OPENAI_API_KEY')

from PIL import Image
import requests
import numpy as np

from langchain.chat_models import ChatOpenAI
from langchain.schema.messages import HumanMessage, SystemMessage

image_path = 'https://i.imgur.com/IxC9lgd.png'

chat = ChatOpenAI(model="gpt-4-vision-preview", max_tokens=512, openai_api_key=openai_api_key)
result = chat.invoke(
    [
        HumanMessage(
            content=[
                {"type": "text", "text": "What is in the picture? Please provide a detailed introduction."},
                {
                    "type": "image_url",
                    "image_url": {
                        "url": image_path,
                        "detail": "auto",
                    },
                },
            ]
        )
    ]
)


print(result.content)

from langchain.prompts import PromptTemplate
from langchain.chat_models import ChatOpenAI
from langchain.schema import StrOutputParser

chain = ChatOpenAI(model="gpt-3.5-turbo", max_tokens=1024)

prompt = PromptTemplate.from_template(
"""
Extract the list and child's name from the text below and return the data in JSON format using the following name:
- "child_name", "wishlist".

{santalist}

"""
)

runnable = prompt | chain | StrOutputParser()

letter = result.content
wishlist = runnable.invoke({"santalist": letter})
print(wishlist)

chain = ChatOpenAI(model="gpt-3.5-turbo", max_tokens=1024)

prompt = PromptTemplate.from_template(
"""

{santalist_json}

From the JSON above, include a new attribute in the JSON called 'weight',
which will calculate the total estimated weight of each item in the list in kilograms.

You will first need to estimate the weight of each item individually.
After that, sum these values to obtain the total weight.
Extract only the numerical value.

"""
)

runnable = prompt | chain | StrOutputParser()

new_wishlist = runnable.invoke({"santalist_json": wishlist})
print(new_wishlist)

from elasticsearch import Elasticsearch
import json

url = f"https://{elastic_user}:{elastic_password}@{elastic_endpoint}:9200"
es = Elasticsearch(url, ca_certs = "./http_ca.crt", verify_certs = True)

es.info() # should return cluster info

# Parse the JSON string
json_string = new_wishlist
data = json.loads(json_string)

# Index name
index_name = "santa_claus_list"

# Index the document
response = es.index(index=index_name, document=data)

# Print the response from Elasticsearch
print(response)

POST /_query?format=txt
{
  "query": """
  FROM santa_claus_list
  | STATS  sum_toy = SUM(weight) BY child_name
  | LIMIT 100
  """
}

# result
    sum_toy    |  child_name   
---------------+---------------
30.5           |Maria
1.5            |Mike
3.0            |Theo
2.5            |Isabella
40.0           |William
30.0           |Olivia