| import streamlit as st |
| import os |
| from langchain.llms import HuggingFaceHub |
| from langchain_huggingface import HuggingFaceEndpoint |
| from langchain_huggingface import HuggingFaceEmbeddings |
| from langchain_huggingface import ChatHuggingFace |
| from langchain.prompts import ChatPromptTemplate |
| from langchain_community.document_loaders import PyPDFLoader |
| from langchain.text_splitter import CharacterTextSplitter,RecursiveCharacterTextSplitter |
| from langchain_community.vectorstores import FAISS,Chroma |
| from langchain_core.output_parsers import StrOutputParser |
| from langchain_core.runnables import RunnablePassthrough |
|
|
| hf_tkn = os.getenv("HF_TOKEN") |
| st.title('Hey Dear! Ask me anything about Kenyan constitution.') |
|
|
| input_text = st.text_area('Write a message...') |
|
|
| pdfloader = PyPDFLoader('Constitution of Kenya 2010.pdf') |
| docs = pdfloader.load() |
|
|
| splitter = RecursiveCharacterTextSplitter(chunk_size=1000,chunk_overlap=0) |
| texts = splitter.split_documents(docs) |
|
|
| embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2") |
| db = FAISS.from_documents(texts,embedding=embeddings) |
|
|
| llm = HuggingFaceEndpoint( |
| repo_id="HuggingFaceH4/zephyr-7b-beta", |
| task="text-generation", |
| max_new_tokens=512, |
| do_sample=False, |
| repetition_penalty=1.03, |
| huggingfacehub_api_token=hf_tkn |
| ) |
|
|
| chat_model = ChatHuggingFace(llm=llm) |
|
|
| prompt = ChatPromptTemplate.from_template(""" |
| Answer the following question based only on the provided context |
| Think step by step before providing a detailed answer |
| <context> |
| {context} |
| </context> |
| Question: {input}""") |
|
|
| retriever = db.as_retriever() |
|
|
| retrieval_chain = ( |
| {"context":retriever,"input":RunnablePassthrough()} |
| | prompt |
| | chat_model |
| | StrOutputParser() |
| ) |
|
|
| def capitalize_first_letter(response): |
| return response[0].upper() + response[1:] if response else response |
|
|
| if st.button('Response'): |
| response = retrieval_chain.invoke(input_text).replace("Based on the provided context, ", "") |
| response = capitalize_first_letter(response) |
| st.success(response) |
|
|
|
|
|
|
