mirror of
https://github.com/hwchase17/langchain
synced 2024-11-18 09:25:54 +00:00
52 lines
1.3 KiB
Python
52 lines
1.3 KiB
Python
from operator import itemgetter
|
|
|
|
from langchain_community.chat_models import ChatOpenAI
|
|
from langchain_core.output_parsers import StrOutputParser
|
|
from langchain_core.prompts import ChatPromptTemplate
|
|
from langchain_core.pydantic_v1 import BaseModel
|
|
from langchain_core.runnables import ConfigurableField, RunnableParallel
|
|
|
|
from neo4j_advanced_rag.retrievers import (
|
|
hypothetic_question_vectorstore,
|
|
parent_vectorstore,
|
|
summary_vectorstore,
|
|
typical_rag,
|
|
)
|
|
|
|
template = """Answer the question based only on the following context:
|
|
{context}
|
|
|
|
Question: {question}
|
|
"""
|
|
prompt = ChatPromptTemplate.from_template(template)
|
|
|
|
model = ChatOpenAI()
|
|
|
|
retriever = typical_rag.as_retriever().configurable_alternatives(
|
|
ConfigurableField(id="strategy"),
|
|
default_key="typical_rag",
|
|
parent_strategy=parent_vectorstore.as_retriever(),
|
|
hypothetical_questions=hypothetic_question_vectorstore.as_retriever(),
|
|
summary_strategy=summary_vectorstore.as_retriever(),
|
|
)
|
|
|
|
chain = (
|
|
RunnableParallel(
|
|
{
|
|
"context": itemgetter("question") | retriever,
|
|
"question": itemgetter("question"),
|
|
}
|
|
)
|
|
| prompt
|
|
| model
|
|
| StrOutputParser()
|
|
)
|
|
|
|
|
|
# Add typing for input
|
|
class Question(BaseModel):
|
|
question: str
|
|
|
|
|
|
chain = chain.with_types(input_type=Question)
|