import gradio as gr from langchain.document_loaders import OnlinePDFLoader from langchain.text_splitter import CharacterTextSplitter from langchain.llms import HuggingFaceHub from langchain.embeddings import HuggingFaceHubEmbeddings from langchain.vectorstores import Chroma from langchain.chains import RetrievalQA import os import tempfile import openai import json import re from langchain.docstore.document import Document from langchain.document_loaders import TextLoader from langchain.text_splitter import CharacterTextSplitter from langchain.embeddings.openai import OpenAIEmbeddings from langchain.vectorstores import FAISS from langchain.chains import RetrievalQA from langchain.llms import OpenAI from langchain.document_loaders import PyPDFLoader from langchain.indexes import VectorstoreIndexCreator import tempfile from langchain.document_loaders import DirectoryLoader os.environ["OPENAI_API_KEY"] = os.environ['OpenApi_Key'] query1=" " limit = 0 def loading_pdf(): return "Loading..." def remove_last_slash_and_characters(input_string): last_slash_index = input_string.rfind('/') if last_slash_index != -1: result_string = input_string[:last_slash_index] else: result_string = input_string return result_string def pdf_changes(pdf_doc, prompt): loader = DirectoryLoader(remove_last_slash_and_characters(pdf_doc.name)) #loader = OnlinePDFLoader(pdf_doc.name) #loader = PyPDFLoader(pdf_doc.name) documents = loader.load() print(len(documents)) #name_filter = "**/*.md" separator = "\n" chunk_size_limit = 1000 max_chunk_overlap = 50 text_splitter = CharacterTextSplitter(separator=separator, chunk_size=chunk_size_limit, chunk_overlap=max_chunk_overlap) split_docs = text_splitter.split_documents(documents) embeddings = OpenAIEmbeddings() vector_store1 = FAISS.from_documents(split_docs, embeddings) from langchain.prompts import ( ChatPromptTemplate, SystemMessagePromptTemplate, HumanMessagePromptTemplate, ) system_template="""You are a helpful chatbot used by the user to chat with pdf documents. Only answer the questions by using information provided in the context provided to you. If there is no relavant context, tell 'Hmm, I'm not sure'."""+prompt+""" ---------------- {summaries}""" messages = [ SystemMessagePromptTemplate.from_template(system_template), HumanMessagePromptTemplate.from_template("{question}") ] prompt2 = ChatPromptTemplate.from_messages(messages) from langchain.chat_models import ChatOpenAI from langchain.chains import RetrievalQAWithSourcesChain global query1 chain_type_kwargs = {"prompt": prompt2} llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0, max_tokens=512) global chain chain = RetrievalQAWithSourcesChain.from_chain_type( llm=llm, chain_type="stuff", retriever=vector_store1.as_retriever(search_kwargs={'k': 2}), return_source_documents=True, chain_type_kwargs=chain_type_kwargs ) return "Ready" def add_text(history, text): history = history + [(text, None)] return history, "" def bot(history): response = infer(history[-1][0]) history[-1][1] = response return history def infer(question): global query1 global limit openai.api_key = os.environ['OpenApi_Key'] prompt_text = question if prompt_text: query1 = query1 + "\nUser: " + prompt_text + "\nBot: " result = chain(query1) query1 = query1 + result['answer'] query1 = openai.ChatCompletion.create( model="gpt-3.5-turbo", messages=[ {"role": "system", "content": "You are provided with chat history and latset conversation between user and bot. Summarise the history and latest conversationin minimum most tokens possible. Do not include greetings in the summary like hi, hello, etc."}, {"role": "user", "content": query1}, ] ) ["choices"][0]["message"]["content"].replace("'", "") return result['answer'] css=""" #col-container { margin-left: auto; margin-right: auto;} """ title = """
Upload a .PDF from your computer, click the "Load PDF" button,
when everything is ready, you can start asking questions about the pdf ;)