|
from langchain.retrievers.multi_query import MultiQueryRetriever |
|
from langchain_openai import ChatOpenAI |
|
from langchain_community.document_transformers import LongContextReorder |
|
from langchain_text_splitters import RecursiveCharacterTextSplitter |
|
from langchain.schema import Document |
|
from langchain_chroma import Chroma |
|
from langchain_openai import OpenAIEmbeddings |
|
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder |
|
from langchain.chains import create_retrieval_chain |
|
from langchain.chains.combine_documents import create_stuff_documents_chain |
|
from langchain_core.messages import HumanMessage |
|
|
|
|
|
|
|
def generate_response(input, history): |
|
|
|
vector_db = Chroma(persist_directory="vector_db", embedding_function=OpenAIEmbeddings()) |
|
|
|
|
|
retriever = vector_db.as_retriever( |
|
search_type="similarity_score_threshold", search_kwargs={"score_threshold": 0.7, "k": 10} |
|
) |
|
|
|
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0.2) |
|
retriever = MultiQueryRetriever.from_llm( |
|
retriever=retriever, llm=llm |
|
) |
|
|
|
unique_docs = retriever.invoke(input) |
|
reordering = LongContextReorder() |
|
reordered_docs = reordering.transform_documents(unique_docs) |
|
|
|
SYSTEM_TEMPLATE = """ |
|
Answer the user's questions based on the below context. |
|
The context will always be relevant to the question. |
|
Use chat history if required to answer the question. |
|
Always be very descriptive, unless stated otherwise. |
|
Explain as if you are a master at this subject. |
|
If you are asked to show with an formula with anexample, then explain the basics first, then ask for specific values to use for the formula. |
|
<context> |
|
{context} |
|
</context> |
|
<chat_history> |
|
{chat_history} |
|
</chat_history> |
|
""" |
|
|
|
question_answering_prompt = ChatPromptTemplate.from_messages( |
|
[ |
|
( |
|
"system", |
|
SYSTEM_TEMPLATE, |
|
), |
|
MessagesPlaceholder(variable_name="messages"), |
|
] |
|
) |
|
|
|
document_chain = create_stuff_documents_chain(llm, question_answering_prompt) |
|
|
|
response = document_chain.invoke( |
|
{ |
|
"context": reordered_docs, |
|
"messages": [ |
|
HumanMessage(content=input) |
|
], |
|
"chat_history": history |
|
} |
|
) |
|
|
|
return response |
|
|
|
|
|
if __name__ == "__main__": |
|
prompt = "How should my MDPE pipe cross a Nallah?" |
|
response = generate_response(prompt, []) |
|
print(response) |
|
|
