docs/how_to/sequence/

[giscus[bot]]

docs/how_to/sequence/

This guide assumes familiarity with the following concepts:

https://python.langchain.com/docs/how_to/sequence/

[KeivanJamali]

Can you please tell me how can I use pipe to make retrieval chain? for example how to write this by "|"?
RetrievalQA.from_chain_type(llm=model,
chain_type="stuff",
retriever=retriever)

[palaemezie]

Hi Keivan,

I don't think you can use the pipe operator, |, in the same way you use the traditional RetrievalQA chain. Here's a generic way of implementing the pipe operator for creating retrieval chains in a more modern way

from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from operator import itemgetter

First, create your prompt template

prompt = ChatPromptTemplate.from_template("""Answer the question based on the following context:

Context: {context}

Question: {question}

Answer:""")

Create the retrieval chain using pipe syntax

retrieval_chain = (
{"context": itemgetter("question") | retriever,
"question": itemgetter("question")}
| prompt
| llm
| StrOutputParser()
)

Use the chain

response = retrieval_chain.invoke({"question": "Your question here"})

Also, here's how I used it in one of the projects I worked on while taking a course on RAG.

generate_qa = (
{"context": itemgetter("sub_question") | retriever,
"sub_question": itemgetter("sub_question"),
"q_a_pairs": itemgetter("q_a_pairs")}
| prompt_qa
| llm
| StrOutputParser()
)