from langchain.text_splitter import CharacterTextSplitter from langchain.vectorstores import FAISS from langchain.chat_models import ChatOpenAI from langchain_openai import AzureChatOpenAI,AzureOpenAIEmbeddings from langchain.memory import ConversationBufferMemory from langchain.chains import ConversationChain from langchain.chains import ConversationalRetrievalChain from langchain.document_loaders import UnstructuredFileLoader from typing import List, Dict, Tuple import gradio as gr import validators import requests import mimetypes import tempfile import os from langchain.chains.question_answering import load_qa_chain from langchain.llms import OpenAI from langchain.prompts import PromptTemplate from langchain.prompts.prompt import PromptTemplate import pandas as pd from langchain_experimental.agents.agent_toolkits import create_csv_agent # from langchain.agents import create_pandas_dataframe_agent # from langchain.agents import ZeroShotAgent, Tool, AgentExecutor from langchain.agents.agent_types import AgentType # from langchain.agents import create_csv_agent from langchain import OpenAI, LLMChain from openai import AzureOpenAI class ChatDocumentQA: def __init__(self) -> None: pass def _get_empty_state(self) -> Dict[str, None]: """Create an empty knowledge base.""" return {"knowledge_base": None} def _extract_text_from_pdfs(self, file_paths: List[str]) -> List[str]: """Extract text content from PDF files. Args: file_paths (List[str]): List of file paths. Returns: List[str]: Extracted text from the PDFs. """ docs = [] loaders = [UnstructuredFileLoader(file_obj, strategy="fast") for file_obj in file_paths] for loader in loaders: docs.extend(loader.load()) return docs def _get_content_from_url(self, urls: str) -> List[str]: """Fetch content from given URLs. Args: urls (str): Comma-separated URLs. Returns: List[str]: List of text content fetched from the URLs. """ file_paths = [] for url in urls.split(','): if validators.url(url): headers = {'user-agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36',} r = requests.get(url, headers=headers) if r.status_code != 200: raise ValueError("Check the url of your file; returned status code %s" % r.status_code) content_type = r.headers.get("content-type") file_extension = mimetypes.guess_extension(content_type) temp_file = tempfile.NamedTemporaryFile(suffix=file_extension, delete=False) temp_file.write(r.content) file_paths.append(temp_file.name) docs = self._extract_text_from_pdfs(file_paths) return docs def _split_text_into_chunks(self, text: str) -> List[str]: """Split text into smaller chunks. Args: text (str): Input text to be split. Returns: List[str]: List of smaller text chunks. """ text_splitter = CharacterTextSplitter(separator="\n", chunk_size=6000, chunk_overlap=0, length_function=len) chunks = text_splitter.split_documents(text) return chunks def _create_vector_store_from_text_chunks(self, text_chunks: List[str]) -> FAISS: """Create a vector store from text chunks. Args: text_chunks (List[str]): List of text chunks. Returns: FAISS: Vector store created from the text chunks. """ embeddings = AzureOpenAIEmbeddings( azure_deployment="openai_embedding", ) return FAISS.from_documents(documents=text_chunks, embedding=embeddings) def _create_conversation_chain(self,vectorstore): _template = """Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language. Chat History: {chat_history} Follow Up Input: {question} Standalone question:""" CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(_template) memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True) # llm = ChatOpenAI(temperature=0) llm=AzureChatOpenAI(azure_deployment = "GPT-3") return ConversationalRetrievalChain.from_llm(llm=llm, retriever=vectorstore.as_retriever(), condense_question_prompt=CONDENSE_QUESTION_PROMPT, memory=memory) def _get_documents_knowledge_base(self, file_paths: List[str]) -> Tuple[str, Dict[str, FAISS]]: """Build knowledge base from uploaded files. Args: file_paths (List[str]): List of file paths. Returns: Tuple[str, Dict]: Tuple containing a status message and the knowledge base. """ file_path = file_paths[0].name file_extension = os.path.splitext(file_path)[1] if file_extension == '.csv': # agent = self.create_agent(file_path) # tools = self.get_agent_tools(agent) # memory,tools,prompt = self.create_memory_for_csv_qa(tools) # agent_chain = self.create_agent_chain_for_csv_qa(memory,tools,prompt) agent_chain = create_csv_agent( AzureChatOpenAI(azure_deployment = "GPT-3"), file_path, verbose=True, agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION, ) return "file uploaded", {"knowledge_base": agent_chain} else: pdf_docs = [file_path.name for file_path in file_paths] raw_text = self._extract_text_from_pdfs(pdf_docs) text_chunks = self._split_text_into_chunks(raw_text) vectorstore = self._create_vector_store_from_text_chunks(text_chunks) return "file uploaded", {"knowledge_base": vectorstore} def _get_urls_knowledge_base(self, urls: str) -> Tuple[str, Dict[str, FAISS]]: """Build knowledge base from URLs. Args: urls (str): Comma-separated URLs. Returns: Tuple[str, Dict]: Tuple containing a status message and the knowledge base. """ webpage_text = self._get_content_from_url(urls) text_chunks = self._split_text_into_chunks(webpage_text) vectorstore = self._create_vector_store_from_text_chunks(text_chunks) return "file uploaded", {"knowledge_base": vectorstore} #************************ # csv qa #************************ def create_agent(self,file_path): agent_chain = create_csv_agent( AzureChatOpenAI(azure_deployment = "GPT-3"), file_path, verbose=True, agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION, ) return agent_chain def get_agent_tools(self,agent): # search = agent tools = [ Tool( name="dataframe qa", func=agent.run, description="useful for when you need to answer questions about table data and dataframe data", ) ] return tools def create_memory_for_csv_qa(self,tools): prefix = """Have a conversation with a human, answering the following questions about table data and dataframe data as best you can. You have access to the following tools:""" suffix = """Begin!" {chat_history} Question: {input} {agent_scratchpad}""" prompt = ZeroShotAgent.create_prompt( tools, prefix=prefix, suffix=suffix, input_variables=["input", "chat_history", "agent_scratchpad"], ) memory = ConversationBufferMemory(memory_key="chat_history",return_messages=True) return memory,tools,prompt def create_agent_chain_for_csv_qa(self,memory,tools,prompt): llm_chain = LLMChain(llm=AzureChatOpenAI(azure_deployment = "GPT-3"), prompt=prompt) agent = ZeroShotAgent(llm_chain=llm_chain, tools=tools, verbose=True) agent_chain = AgentExecutor.from_agent_and_tools( agent=agent, tools=tools, verbose=True, memory=memory ) return agent_chain def _get_response(self, message: str, chat_history: List[Tuple[str, str]], state: Dict[str, FAISS],file_paths) -> Tuple[str, List[Tuple[str, str]]]: """Get a response from the chatbot. Args: message (str): User's message/question. chat_history (List[Tuple[str, str]]): List of chat history as tuples of (user_message, bot_response). state (dict): State containing the knowledge base. Returns: Tuple[str, List[Tuple[str, str]]]: Tuple containing a status message and updated chat history. """ try: if file_paths: file_path = file_paths[0].name file_extension = os.path.splitext(file_path)[1] if file_extension == '.csv': agent_chain = state["knowledge_base"] response = agent_chain.run(input = message) chat_history.append((message, response)) return "", chat_history else: vectorstore = state["knowledge_base"] chat = self._create_conversation_chain(vectorstore) print("chat_history",chat_history) response = chat({"question": message,"chat_history": chat_history}) chat_history.append((message, response["answer"])) return "", chat_history except: chat_history.append((message, "Please Upload Document or URL")) return "", chat_history def gradio_interface(self) -> None: """Create a Gradio interface for the chatbot.""" with gr.Blocks(css = "style.css" ,theme="freddyaboulton/test-blue") as demo: gr.HTML("""


Virtual Assistant Chatbot

""") state = gr.State(self._get_empty_state()) chatbot = gr.Chatbot() with gr.Row(): with gr.Column(scale=0.85): msg = gr.Textbox(label="Question") with gr.Column(scale=0.15): file_output = gr.Textbox(label="File Status") with gr.Row(): with gr.Column(scale=0.85): clear = gr.ClearButton([msg, chatbot]) with gr.Column(scale=0.15): upload_button = gr.UploadButton( "Browse File", file_types=[".txt", ".pdf", ".doc", ".docx", ".csv"], file_count="multiple", variant="primary" ) with gr.Row(): with gr.Column(scale=1): input_url = gr.Textbox(label="urls") input_url.submit(self._get_urls_knowledge_base, input_url, [file_output, state]) upload_button.upload(self._get_documents_knowledge_base, upload_button, [file_output, state]) msg.submit(self._get_response, [msg, chatbot, state,upload_button], [msg, chatbot]) demo.launch() if __name__ == "__main__": chatdocumentqa = ChatDocumentQA() chatdocumentqa.gradio_interface()