app.py

# from typing import Any, Coroutine
import openai
import os
from langchain.vectorstores import Chroma
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.text_splitter import CharacterTextSplitter
from langchain.chat_models import AzureChatOpenAI
from langchain.document_loaders import DirectoryLoader
from langchain.chains import RetrievalQA  
from langchain.vectorstores import Pinecone
from langchain.agents import initialize_agent
from langchain.agents import AgentType
from langchain.agents import Tool
# from langchain.agents import load_tools
from langchain.tools import BaseTool
from langchain.tools import DuckDuckGoSearchRun
from langchain.utilities import WikipediaAPIWrapper
from langchain.python import PythonREPL
from langchain.chains import LLMMathChain
from langchain.memory import ConversationBufferMemory
from langchain.agents import ZeroShotAgent, AgentExecutor
from langchain import LLMChain
import azure.cognitiveservices.speech as speechsdk
import requests

import pinecone      
from pinecone.core.client.configuration import Configuration as OpenApiConfiguration
import gradio as gr
import time

import glob
from typing import List
from multiprocessing import Pool
from tqdm import tqdm

from langchain.document_loaders import (
    CSVLoader,
    EverNoteLoader,
    PyMuPDFLoader,
    TextLoader,
    UnstructuredEmailLoader,
    UnstructuredEPubLoader,
    UnstructuredHTMLLoader,
    UnstructuredMarkdownLoader,
    UnstructuredODTLoader,
    UnstructuredPowerPointLoader,
    UnstructuredWordDocumentLoader,
)
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.docstore.document import Document


memory = ConversationBufferMemory(memory_key="chat_history")
# Custom document loaders
class MyElmLoader(UnstructuredEmailLoader):
    """Wrapper to fallback to text/plain when default does not work"""

    def load(self) -> List[Document]:
        """Wrapper adding fallback for elm without html"""
        try:
            try:
                doc = UnstructuredEmailLoader.load(self)
            except ValueError as e:
                if 'text/html content not found in email' in str(e):
                    # Try plain text
                    self.unstructured_kwargs["content_source"]="text/plain"
                    doc = UnstructuredEmailLoader.load(self)
                else:
                    raise
        except Exception as e:
            # Add file_path to exception message
            raise type(e)(f"{self.file_path}: {e}") from e

        return doc
    
LOADER_MAPPING = {
    ".csv": (CSVLoader, {}),
    # ".docx": (Docx2txtLoader, {}),
    ".doc": (UnstructuredWordDocumentLoader, {}),
    ".docx": (UnstructuredWordDocumentLoader, {}),
    ".enex": (EverNoteLoader, {}),
    ".eml": (MyElmLoader, {}),
    ".epub": (UnstructuredEPubLoader, {}),
    ".html": (UnstructuredHTMLLoader, {}),
    ".md": (UnstructuredMarkdownLoader, {}),
    ".odt": (UnstructuredODTLoader, {}),
    ".pdf": (PyMuPDFLoader, {}),
    ".ppt": (UnstructuredPowerPointLoader, {}),
    ".pptx": (UnstructuredPowerPointLoader, {}),
    ".txt": (TextLoader, {"encoding": "utf8"}),
    # Add more mappings for other file extensions and loaders as needed
}

source_directory = 'Upload Files'
global file_list_loaded
file_list_loaded = ''
chunk_size = 500
chunk_overlap = 300

global Audio_output
Audio_output = []


def load_single_document(file_path: str) -> List[Document]:
    ext = "." + file_path.rsplit(".", 1)[-1]
    if ext in LOADER_MAPPING:
        loader_class, loader_args = LOADER_MAPPING[ext]
        loader = loader_class(file_path, **loader_args)
        return loader.load()

    raise ValueError(f"Unsupported file extension '{ext}'")


def load_documents(source_dir: str, ignored_files: List[str] = []) -> List[Document]:
    """
    Loads all documents from the source documents directory, ignoring specified files
    """
    all_files = []
    for ext in LOADER_MAPPING:
        all_files.extend(
            glob.glob(os.path.join(source_dir, f"**/*{ext}"), recursive=True)
        )
    filtered_files = [file_path for file_path in all_files if file_path not in ignored_files]

    with Pool(processes=os.cpu_count()) as pool:
        results = []
        with tqdm(total=len(filtered_files), desc='Loading new documents', ncols=80) as pbar:
            for i, docs in enumerate(pool.imap_unordered(load_single_document, filtered_files)):
                results.extend(docs)
                pbar.update()

    return results

def load_documents_2(all_files: List[str] = [], ignored_files: List[str] = []) -> List[Document]:
    """
    Loads all documents from the source documents directory, ignoring specified files
    """
    # all_files = []
    # for ext in LOADER_MAPPING:
    #     all_files.extend(
    #         glob.glob(os.path.join(source_dir, f"**/*{ext}"), recursive=True)
    #     )
    filtered_files = [file_path for file_path in all_files if file_path not in ignored_files]


    results = []
    with tqdm(total=len(filtered_files), desc='Loading new documents', ncols=80) as pbar:
        for file in filtered_files:
            docs = load_single_document(file)
            results.extend(docs)
            pbar.update()

    return results


def process_documents(ignored_files: List[str] = []) -> List[Document]:
    """
    Load documents and split in chunks
    """
    print(f"Loading documents from {source_directory}")
    documents = load_documents(source_directory, ignored_files)
    if not documents:
        print("No new documents to load")
        exit(0)
    print(f"Loaded {len(documents)} new documents from {source_directory}")
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
    texts = text_splitter.split_documents(documents)
    print(f"Split into {len(texts)} chunks of text (max. {chunk_size} tokens each)")
    return texts

def process_documents_2(ignored_files: List[str] = []) -> List[Document]:
    """
    Load documents and split in chunks
    """
    global file_list_loaded
    print(f"Loading documents from {source_directory}")
    print("File Path to start processing:", file_list_loaded)
    documents = load_documents_2(file_list_loaded, ignored_files)
    if not documents:
        print("No new documents to load")
        exit(0)
    print(f"Loaded {len(documents)} new documents from {source_directory}")
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
    texts = text_splitter.split_documents(documents)
    print(f"Split into {len(texts)} chunks of text (max. {chunk_size} tokens each)")
    return texts

def UpdateDb():
    global vectordb_p
    # pinecone.Index(index_name).delete(delete_all=True, namespace='')
    # collection = vectordb_p.get()
    # split_docs = process_documents([metadata['source'] for metadata in collection['metadatas']])
    # split_docs = process_documents()
    split_docs = process_documents_2()
    tt = len(split_docs)
    print(split_docs[tt-1])
    print(f"Creating embeddings. May take some minutes...")
    vectordb_p = Pinecone.from_documents(split_docs, embeddings, index_name = "stla-baby")
    print("Pinecone Updated Done")
    print(index.describe_index_stats())


class DB_Search(BaseTool):
    name = "Vector Database Search"
    description = "This is the internal database to search information firstly. If information is found, it is trustful."
    def _run(self, query: str) -> str:
        response, source = QAQuery_p(query)
        # response = "test db_search feedback"
        return response

    def _arun(self, query: str):
        raise NotImplementedError("N/A")

def Text2Sound(text):

    speech_config = speechsdk.SpeechConfig(subscription=os.getenv('SPEECH_KEY'), region=os.getenv('SPEECH_REGION'))
    audio_config = speechsdk.audio.AudioOutputConfig(use_default_speaker=True)
    speech_config.speech_synthesis_voice_name='en-US-JennyNeural'
    # speech_synthesizer = ""
    speech_synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=audio_config)
    speech_synthesis_result = speech_synthesizer.speak_text_async(text).get()
    # if speech_synthesis_result.reason == speechsdk.ResultReason.SynthesizingAudioCompleted:
    #     print("Speech synthesized for text [{}]".format(text))
    # elif speech_synthesis_result.reason == speechsdk.ResultReason.Canceled:
    #     cancellation_details = speech_synthesis_result.cancellation_details
    #     print("Speech synthesis canceled: {}".format(cancellation_details.reason))
    #     if cancellation_details.reason == speechsdk.CancellationReason.Error:
    #         if cancellation_details.error_details:
    #             print("Error details: {}".format(cancellation_details.error_details))
    #             print("Did you set the speech resource key and region values?")
    print("test")
    return speech_synthesis_result
    pass



def get_azure_access_token():
    azure_key = os.environ.get("SPEECH_KEY")
    try:
        response = requests.post(
            "https://eastus.api.cognitive.microsoft.com/sts/v1.0/issuetoken",
            headers={
                "Ocp-Apim-Subscription-Key": azure_key
            }
        )
        response.raise_for_status()
    except requests.exceptions.RequestException as e:
        print(f"Error: {e}")
        return None
    # print (response.text)
    return response.text


def text_to_speech_2(text):
    global Audio_output
    access_token = get_azure_access_token()
    voice_name='en-US-AriaNeural'
    if not access_token:
        return None

    try:
        response = requests.post(
            "https://eastus.tts.speech.microsoft.com/cognitiveservices/v1",
            headers={
                "Authorization": f"Bearer {access_token}",
                "Content-Type": "application/ssml+xml",
                "X-MICROSOFT-OutputFormat": "riff-24khz-16bit-mono-pcm",
                "User-Agent": "TextToSpeechApp",
            },
            data=f"""
                <speak version='1.0' xml:lang='en-US'>
                <voice name='{voice_name}'>
                    {text}
                </voice>
                </speak>
            """,
        )
        response.raise_for_status()
        timestr = time.strftime("%Y%m%d-%H%M")
        with open('sample-' + timestr + '.wav', 'wb') as audio:
                audio.write(response.content)
        print ("File Name  ", audio.name)
        # print (audio)
        Audio_output.append(audio.name)
        return audio.name
    except requests.exceptions.RequestException as e:
        print(f"Error: {e}")
        return None
    
Text2Sound_tool = Tool(
    name = "Text To Sound REST API",
    # func = Text2Sound,
    func = text_to_speech_2,
    description = "Useful when you need to convert text into sound file."
)

Wikipedia = WikipediaAPIWrapper()
Netsearch = DuckDuckGoSearchRun()
Python_REPL = PythonREPL()

wikipedia_tool = Tool(
    name = "Wikipedia Search",
    func = Wikipedia.run,
    description = "Useful to search a topic, country or person when there is no availble information in vector database"
)

duckduckgo_tool = Tool(
    name = "Duckduckgo Internet Search",
    func = Netsearch.run,
    description = "Useful to search information in internet when it is not available in other tools"    
)

python_tool = Tool(
    name = "Python REPL",
    func = Python_REPL.run,
    description = "Useful when you need python to answer questions. You should input python code."    
)

# tools = [DB_Search(), wikipedia_tool, duckduckgo_tool, python_tool]


os.environ["OPENAI_API_TYPE"] = "azure"
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
os.environ["OPENAI_API_BASE"] = os.getenv("OPENAI_API_BASE")
os.environ["OPENAI_API_VERSION"] = "2023-05-15"
username = os.getenv("username")
password = os.getenv("password")
SysLock = os.getenv("SysLock") # 0=unlock 1=lock

# deployment_name="Chattester"

chat = AzureChatOpenAI(
    deployment_name=os.getenv("deployment_name"),
    temperature=0,
)
llm = chat

llm_math = LLMMathChain.from_llm(llm)

math_tool = Tool(
    name ='Calculator',
    func = llm_math.run,
    description ='Useful for when you need to answer questions about math.'
)

tools = [DB_Search(), duckduckgo_tool, wikipedia_tool, python_tool, math_tool, Text2Sound_tool]

# tools = load_tools(["Vector Database Search","Wikipedia Search","Python REPL","llm-math"], llm=llm)

embeddings = OpenAIEmbeddings(deployment="model_embedding", chunk_size=15)


pinecone.init(      
	api_key = os.getenv("pinecone_api_key"),      
	environment='asia-southeast1-gcp-free',
    # openapi_config=openapi_config      
)
index_name = 'stla-baby'     
index = pinecone.Index(index_name)
# index.delete(delete_all=True, namespace='')
# print(pinecone.whoami())
# print(index.describe_index_stats())

"""
Answer the following questions as best you can with details. 
You can always use tools to convert text to sound.
You must always check internal vector database first and try to answer the question based on the information in internal vector database only.
Only when there is no information available from vector database, you can search information by using other tools.
You have access to the following tools:

Vector Database Search: This is the internal database to search information firstly. If information is found, it is trustful.
Duckduckgo Internet Search: Useful to search information in internet when it is not available in other tools.
Wikipedia Search: Useful to search a topic, country or person when there is no availble information in vector database
Python REPL: Useful when you need python to answer questions. You should input python code.
Calculator: Useful for when you need to answer questions about math.
Text To Sound: Useful when you need to convert text into sound file."""


PREFIX = """Answer the following questions as best you can with detail information and explanation. 
You can always use tools to convert text to sound.
You must always check internal vector database first and try to answer the question based on the information found in internal vector database only.
Only when there is no information available from vector database, you can search information by using other tools.
You have access to the following tools:"""

FORMAT_INSTRUCTIONS = """Use the following format:

Question: the input question you must answer
Thought: you should always think about what to do
Action: the action to take, should be one of [Vector Database Search, Duckduckgo Internet Search, Python REPL, Calculator]
Action Input: the input to the action
Observation: the result of the action
... (this Thought/Action/Action Input/Observation can repeat N times)
Thought: I now know the final answer
Final Answer: the final answer to the original input question"""

SUFFIX = """Begin!

Request: {input}
Thought:{agent_scratchpad}"""

SUFFIX2 = """Begin!

{chat_history}
Question: {input}
Thought:{agent_scratchpad}"""


prompt = ZeroShotAgent.create_prompt(
    tools, 
    prefix=PREFIX,
    suffix=SUFFIX2,
    format_instructions=FORMAT_INSTRUCTIONS, 
    input_variables=["input", "chat_history", "agent_scratchpad"]
)

input_variables=["input", "chat_history", "agent_scratchpad"]

agent2 = initialize_agent(tools, llm, 
                        #  agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
                         agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
                         verbose = True,
                         handle_parsing_errors = True,
                         max_iterations = int(os.getenv("max_iterations")),
                         early_stopping_method="generate",
                         agent_kwargs={
                            'prefix': PREFIX,
                            'format_instructions': FORMAT_INSTRUCTIONS,
                            'suffix': SUFFIX,
                            # 'input_variables': input_variables,
                         },
                        #  input_variables = input_variables,
                        #  agent_kwargs={
                        #     'prompt': prompt,
                        #  }
                         
                        )

llm_chain = LLMChain(llm=llm, prompt=prompt)
agent_core = ZeroShotAgent(llm_chain=llm_chain, tools=tools, verbose=True)
agent = AgentExecutor.from_agent_and_tools(
    agent=agent_core, 
    tools=tools, 
    verbose=True, 
    memory=memory,
    handle_parsing_errors = True,
    max_iterations = int(os.getenv("max_iterations")),
    early_stopping_method="generate",
    )

print(agent.agent.llm_chain.prompt.template)
# print(agent.agent.llm_chain.prompt)

global vectordb
vectordb = Chroma(persist_directory='db', embedding_function=embeddings)
global vectordb_p
vectordb_p = Pinecone.from_existing_index(index_name, embeddings)

# loader = DirectoryLoader('./documents', glob='**/*.txt')
# documents = loader.load()
# text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=200)
# split_docs = text_splitter.split_documents(documents)
# print(split_docs)
# vectordb = Chroma.from_documents(split_docs, embeddings, persist_directory='db')



# question = "what is LCDV ?"
# rr = vectordb.similarity_search(query=question, k=4)
# vectordb.similarity_search(question)
# print(type(rr))
# print(rr)
def chathmi(message, history):
    # response = "I don't know"
    # print(message)
    response, source = QAQuery_p(message)
    time.sleep(0.3)
    print(history)
    yield response
    # yield history

def chathmi2(message, history):
    global Audio_output
    try:
        output = agent.run(message)
        time.sleep(0.3)
        response = output
        yield response
        if len(Audio_output) > 0:
            # time.sleep(0.5)        
            yield Audio_output
            Audio_output = []
        print("History: ", history)
        print("-" * 20)
        print("-" * 20)
    except Exception as e:
        print("error:", e)

    # yield history
# chatbot = gr.Chatbot().style(color_map =("blue", "pink"))
# chatbot = gr.Chatbot(color_map =("blue", "pink"))

def func_upload_file(files, chat_history):
    global file_list_loaded
    file_list_loaded = []
    for unit in files:
        file_list_loaded.append(unit.name)
    # file_list_loaded = files
    print(file_list_loaded)
    # print(chat_history)
    # test_msg = ["Request Upload File into DB", "Operation Ongoing...."]
    
    # chat_history.append(test_msg)
    for file in files:
        chat_history = chat_history + [((file.name,), None)]
    yield chat_history
    if os.getenv("SYS_Upload_Enable") == "1":
        UpdateDb()
    test_msg = ["Request Upload File into DB", "Operation Finished"]
    chat_history.append(test_msg)
    yield chat_history

with gr.Blocks() as demo:
    # gr.Markdown("Start typing below and then click **SUBMIT** to see the output.")
    main = gr.ChatInterface(
        chathmi2,
        title="STLA BABY - YOUR FRIENDLY GUIDE",
        description= "v0.3: Powered by MECH Core Team",   
    )
    upload_button = gr.UploadButton("Upload File", file_count="multiple")
    upload_button.upload(func_upload_file, [upload_button, main.chatbot], main.chatbot)
    
# demo = gr.Interface(
#     chathmi,
#     ["text", "state"],
#     [chatbot, "state"],
#     allow_flagging="never",
# )

def CreatDb_P():
    global vectordb_p
    index_name = 'stla-baby'
    loader = DirectoryLoader('./documents', glob='**/*.txt')
    documents = loader.load()
    text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=200)
    split_docs = text_splitter.split_documents(documents)
    print(split_docs)
    pinecone.Index(index_name).delete(delete_all=True, namespace='')
    vectordb_p = Pinecone.from_documents(split_docs, embeddings, index_name = "stla-baby")
    print("Pinecone Updated Done")
    print(index.describe_index_stats())

def QAQuery_p(question: str):
    global vectordb_p
    # vectordb = Chroma(persist_directory='db', embedding_function=embeddings)
    retriever = vectordb_p.as_retriever()
    retriever.search_kwargs['k'] = int(os.getenv("search_kwargs_k"))
    # retriever.search_kwargs['fetch_k'] = 100

    qa = RetrievalQA.from_chain_type(llm=chat, chain_type="stuff", 
                                     retriever=retriever, return_source_documents = True,
                                     verbose = True)
    # qa = VectorDBQA.from_chain_type(llm=chat, chain_type="stuff", vectorstore=vectordb, return_source_documents=True)
    # res = qa.run(question)
    res = qa({"query": question})
    
    print("-" * 20)
    print("Question:", question)
    # print("Answer:", res)
    print("Answer:", res['result'])
    print("-" * 20)
    print("Source:", res['source_documents'])
    response = res['result']
    # response = res['source_documents']
    source = res['source_documents']
    return response, source

def CreatDb():
    global vectordb
    loader = DirectoryLoader('./documents', glob='**/*.txt')
    documents = loader.load()
    text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=200)
    split_docs = text_splitter.split_documents(documents)
    print(split_docs)
    vectordb = Chroma.from_documents(split_docs, embeddings, persist_directory='db')
    vectordb.persist()

def QAQuery(question: str):
    global vectordb
    # vectordb = Chroma(persist_directory='db', embedding_function=embeddings)
    retriever = vectordb.as_retriever()
    retriever.search_kwargs['k'] = 3
    # retriever.search_kwargs['fetch_k'] = 100

    qa = RetrievalQA.from_chain_type(llm=chat, chain_type="stuff", retriever=retriever, return_source_documents = True)
    # qa = VectorDBQA.from_chain_type(llm=chat, chain_type="stuff", vectorstore=vectordb, return_source_documents=True)
    # res = qa.run(question)
    res = qa({"query": question})
    
    print("-" * 20)
    print("Question:", question)
    # print("Answer:", res)
    print("Answer:", res['result'])
    print("-" * 20)
    print("Source:", res['source_documents'])
    response = res['result']
    return response

# Used to complete content
def completeText(Text): 
    deployment_id="Chattester"
    prompt = Text
    completion = openai.Completion.create(deployment_id=deployment_id,
                                        prompt=prompt, temperature=0)                              
    print(f"{prompt}{completion['choices'][0]['text']}.")

# Used to chat
def chatText(Text): 
    deployment_id="Chattester"
    conversation = [{"role": "system", "content": "You are a helpful assistant."}]
    user_input = Text
    conversation.append({"role": "user", "content": user_input})
    response = openai.ChatCompletion.create(messages=conversation,
        deployment_id="Chattester")
    print("\n" + response["choices"][0]["message"]["content"] + "\n")

if __name__ == '__main__':
    # chatText("what is AI?")
    # CreatDb()
    # QAQuery("what is COFOR ?")
    # CreatDb_P()
    # QAQuery_p("what is GST ?")
    if SysLock == "1":
        demo.queue().launch(auth=(username, password), server_name="0.0.0.0", server_port=7860)
    else:
        demo.queue().launch(server_name="0.0.0.0", server_port=7860)
    pass