import gradio as gr

from langchain.document_loaders import OnlinePDFLoader

from langchain.text_splitter import CharacterTextSplitter
text_splitter = CharacterTextSplitter(chunk_size=350, chunk_overlap=0)

from langchain.llms import HuggingFaceHub
flan_ul2 = HuggingFaceHub(repo_id="google/flan-ul2", model_kwargs={"temperature":0.1, "max_new_tokens":300})

from langchain.embeddings import HuggingFaceHubEmbeddings
embeddings = HuggingFaceHubEmbeddings()

from langchain.vectorstores import Chroma

from langchain.chains import RetrievalQA

def infer(pdf_doc):
    loader = OnlinePDFLoader(pdf_doc)
    documents = loader.load()
    texts = text_splitter.split_documents(documents)
    db = Chroma.from_documents(texts, embeddings)
    retriever = db.as_retriever()
    qa = RetrievalQA.from_chain_type(llm=flan_ul2, chain_type="stuff", retriever=retriever, return_source_documents=True)
    query = "What is the title of this paper?"
    result = qa({"query": query})

    return result

gr.Interface(fn=infer, inputs=[gr.Textbox(value="https://arxiv.org/pdf/2304.03757.pdf")], outputs=[gr.Textbox()]).launch()