ChatPromptTemplate和AI Message的用法

ChatPromptTemplate的用法

用法1：

from langchain.chains import LLMChain
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain.chains import LLMMathChain

prompt= ChatPromptTemplate.from_template("tell me the weather of {topic}")
str = prompt.format(topic="shenzhen")
print(str)
1
2
3
4
5
6
7
8
9
10

打印出：

Human: tell me the weather of shenzhen
1

最终和llm一起使用：

import ChatGLM
from langchain.chains import LLMChain
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate

from langchain_community.tools.tavily_search import TavilySearchResults
from langchain.chains import LLMMathChain


prompt = ChatPromptTemplate.from_template("who is {name}")
# str = prompt.format(name="Bill Gates")
# print(str)
llm = ChatGLM.ChatGLM_LLM()
output_parser = StrOutputParser()
chain05 = prompt| llm | output_parser
print(chain05.invoke({"name": "Bill Gates"}))
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

用法2：

import ChatGLM
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
                ("system", "You are a helpful AI bot. Your name is {name}."),
                ("human", "Hello, how are you doing?"),
                ("ai", "I'm doing well, thanks!"),
                ("human", "{user_input}"),
            ])

llm = ChatGLM.ChatGLM_LLM()
output_parser = StrOutputParser()
chain05 = prompt| llm | output_parser
print(chain05.invoke({"name": "Bob","user_input": "What is your name"}))

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

也可以这样：

import ChatGLM
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate

llm = ChatGLM.ChatGLM_LLM()

prompt = ChatPromptTemplate.from_messages([
                ("system", "You are a helpful AI bot. Your name is {name}."),
                ("human", "Hello, how are you doing?"),
                ("ai", "I'm doing well, thanks!"),
                ("human", "{user_input}"),
            ])


# a = prompt.format_prompt({name="Bob"})

a = prompt.format_prompt(name="Bob",user_input="What is your name") 
print(a)
print(llm.invoke(a))
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

以下也是一个例子：

import gradio as gr
from langchain_core.prompts import ChatPromptTemplate
from LLMs import myllm
from langchain_core.output_parsers import StrOutputParser

llm = myllm()

parser = StrOutputParser()
template = """{question}"""
prompt = ChatPromptTemplate.from_template(template)
chain = prompt | llm | parser

def greet3(name):
    return chain.invoke({"question": name})

def alternatingly_agree(message, history):
   return greet3(message)

gr.ChatInterface(alternatingly_agree).launch(server_name="0.0.0.0",share=False)


1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

参考： https://python.langchain.com/docs/modules/model_io/prompts/quick_start
https://python.langchain.com/docs/modules/model_io/prompts/composition