https:\/\/gist.github.com\/vndee\/7776debe50b5e6c2b174add8646a4625<\/a><\/p>\nCopy the following source code into a Python file with the name rag.py<\/code>.<\/p>\n
\n\n
from langchain.vectorstores import Chroma<\/code><\/div>\nfrom langchain.chat_models import ChatOllama<\/code><\/div>\nfrom langchain.embeddings import FastEmbedEmbeddings<\/code><\/div>\nfrom langchain.schema.output_parser import StrOutputParser<\/code><\/div>\nfrom langchain.document_loaders import PyPDFLoader<\/code><\/div>\nfrom langchain.text_splitter import RecursiveCharacterTextSplitter<\/code><\/div>\nfrom langchain.schema.runnable import RunnablePassthrough<\/code><\/div>\nfrom langchain.prompts import PromptTemplate<\/code><\/div>\nfrom langchain.vectorstores.utils import filter_complex_metadata<\/code><\/div>\nclass ChatPDF:<\/code><\/div>\n\u00a0 \u00a0 vector_store = None<\/code><\/div>\n\u00a0 \u00a0 retriever = None<\/code><\/div>\n\u00a0 \u00a0 chain = None<\/code><\/div>\n\u00a0 \u00a0 def __init__(self):<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 self.model = ChatOllama(model=\"mistral\")<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 self.text_splitter = RecursiveCharacterTextSplitter(chunk_size=1024, chunk_overlap=100)<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 self.prompt = PromptTemplate.from_template(<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \"\"\"<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 <s> [INST] You are an assistant for question-answering tasks. Use the following pieces of retrieved context<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 to answer the question. If you don't know the answer, just say that you don't know. Use three sentences<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0maximum and keep the answer concise. [\/INST] <\/s><\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 [INST] Question: {question}<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 Context: {context}<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 Answer: [\/INST]<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \"\"\"<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 )<\/code><\/div>\n\u00a0 \u00a0 def ingest(self, pdf_file_path: str):<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 docs = PyPDFLoader(file_path=pdf_file_path).load()<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 chunks = self.text_splitter.split_documents(docs)<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 chunks = filter_complex_metadata(chunks)<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 vector_store = Chroma.from_documents(documents=chunks, embedding=FastEmbedEmbeddings())<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 self.retriever = vector_store.as_retriever(<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 search_type=\"similarity_score_threshold\",<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 search_kwargs={<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \"k\": 3,<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \"score_threshold\": 0.5,<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 },<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 )<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 self.chain = ({\"context\": self.retriever, \"question\": RunnablePassthrough()}<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 | self.prompt<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 | self.model<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 | StrOutputParser())<\/code><\/div>\n\u00a0 \u00a0 def ask(self, query: str):<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 if not self.chain:<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 return \"Please, add a PDF document first.\"<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 return self.chain.invoke(query)<\/code><\/div>\n\u00a0 \u00a0 def clear(self):<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 self.vector_store = None<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 self.retriever = None<\/code><\/div>\n\u00a0 \u00a0 \u00a0 \u00a0 self.chain = None<\/code><\/div>\n<\/div>\n<\/div>\n
\n
\n<\/div>\n
Now you need the second file that contains the user interface and includes the
rag.py<\/code> file. I also copied this from the following project on GitHub.<\/div>\n
![\"RAG](\"https:\/\/ai-box.eu\/wp-content\/uploads\/2024\/02\/RAG_Chat_app_architecture-1024x541.png\")
<\/a>