app.py

#from langchain_community.document_loaders import PyPDFLoader

from datasets import load_dataset
dataset = load_dataset("Namitg02/Test")
print(dataset)

from langchain.docstore.document import Document as LangchainDocument

#RAW_KNOWLEDGE_BASE = [LangchainDocument(page_content=["dataset"])]

from langchain.text_splitter import RecursiveCharacterTextSplitter
splitter = RecursiveCharacterTextSplitter(chunk_size=100, chunk_overlap=15,separators=["\n\n", "\n", " ", ""])
#docs = splitter.split_documents(RAW_KNOWLEDGE_BASE)
docs = splitter.create_documents(str(dataset))


from langchain_community.embeddings import HuggingFaceEmbeddings
embedding_model = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
# embeddings = embedding_model.encode(docs)


from langchain_community.vectorstores import Chroma
persist_directory = 'docs/chroma/'

vectordb = Chroma.from_documents(
    documents=docs,
    embedding=embedding_model,
    persist_directory=persist_directory
)

retriever = vectordb.as_retriever()

#docs_ss = vectordb.similarity_search(question,k=3)


#qa_chain = RetrievalQA.from_chain_type(
#    models/HuggingFaceH4/zephyr-7b-beta,
#    retriever=vectordb.as_retriever()
#)

from transformers import pipeline
from transformers import BitsAndBytesConfig
from transformers import AutoTokenizer, AutoModelForCausalLM


READER_MODEL = "HuggingFaceH4/zephyr-7b-beta"
bnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4")
model = AutoModelForCausalLM.from_pretrained(READER_MODEL,quantization_config=bnb_config)
tokenizer = AutoTokenizer.from_pretrained(READER_MODEL)

from langchain.llms import HuggingFacePipeline
from langchain.prompts import PromptTemplate
from transformers import pipeline
from langchain_core.output_parsers import StrOutputParser


text_generation_pipeline = pipeline(
    model=model,
    tokenizer=tokenizer,
    task="text-generation",
    temperature=0.2,
    do_sample=True,
    repetition_penalty=1.1,
    return_full_text=True,
    max_new_tokens=100,
)

llm = HuggingFacePipeline(pipeline=text_generation_pipeline)


#from langchain_community.output_parsers.rail_parser import GuardrailsOutputParser

prompt_template = """
<|system|>
Answer the question based on your knowledge. Use the following context to help:

{context}

</s>
<|user|>
{question}
</s>
<|assistant|>

 """

QA_CHAIN_PROMPT = PromptTemplate(
    input_variables=["context", "question"],
    template=prompt_template,
)


llm_chain = QA_CHAIN_PROMPT | llm | StrOutputParser()


from langchain_core.runnables import RunnablePassthrough


retriever=vectordb.as_retriever()

rag_chain = {"context": retriever, "question": RunnablePassthrough()} | llm_chain


#from langchain.chains import ConversationalRetrievalChain
#from langchain.memory import ConversationBufferMemory
#memory = ConversationBufferMemory(
#    memory_key="chat_history",
#    return_messages=True
#)

question = "Can I reverse Diabetes?"
print("template")

#qa = ConversationalRetrievalChain.from_llm(llm=READER_MODEL,retriever=retriever,memory=memory)

import gradio as gr
gr.load("READER_MODEL").launch()

#result = ({"query": question})
#print("qa")