一、将其他chain的输入作为新chain的输出，三种方式
1、采用连接符"|"，推荐
2、采用lamba表达式输入
3、采用pipe方法
from langchain_community.chat_models import ChatOllama
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnableParallel
prompt=ChatPromptTemplate.from_template("tell me a joke about {topic}"
)
model=ChatOllama(model='qwen:7b')
chain=prompt|model|StrOutputParser()
##批量
# res=chain.batch([{'topic':'bear'},{'topic':'chair'}])
##chain的连接，本例子通过一个chain分析模型的输出结果
analysis_promt=ChatPromptTemplate.from_template("is this a funcy joke?{joke}"
)
###方式1
composed_chian={"joke":chain}|analysis_promt|model|StrOutputParser()
res=composed_chian.invoke({'topic':"bear"})
###方式2
composed_chian_with_lamba=(chain|(lambda x:{"joke":x})|analysis_promt|model|StrOutputParser()
)
res=composed_chian_with_lamba.invoke({'topic':"bear"})
###方式3
composed_chain_with_pipe=(RunnableParallel({'joke':chain}).pipe(analysis_promt).pipe(model).pipe(StrOutputParser())
)
res=composed_chain_with_pipe.invoke({'topic':'bear'})
print(res)
二、RunnableParallel
并行,每个值都是用RunnableParallel的整体输入调用的,使前一个输出格式匹配下一个输入
from langchain_community.vectorstores import FAISS
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough,RunnableParallel
from langchain_community.chat_models import ChatOllama
from langchain_community.embeddings import OllamaEmbeddings
vectorstore=FAISS.from_texts(['harrison worked at kensho'],embedding=OllamaEmbeddings(model='qwen:7b')
)
retriever=vectorstore.as_retriever()
template='''
Answer the question based only on the following context:{context}
Question:{question}
'''
prompt=ChatPromptTemplate.from_template(template)
model=ChatOllama(model='qwen:7b')
retrieval_chain=({'context':retriever,"question":RunnablePassthrough()}##4种等价# RunnableParallel({"context":retriever,"question":RunnablePassthrough()})# RunnableParallel(context=retriever,question=RunnablePassthrough())# {"context":retriever,"question":itemgetter("question")}|prompt|model|StrOutputParser()
)
res=retrieval_chain.invoke({"question":"where did harrison work"})
print(res)
