app.py

# Importing Necessary Libraries
from langchain.chat_models import ChatOpenAI
from langchain.document_loaders import CSVLoader
from langchain.embeddings import OpenAIEmbeddings
from langchain.prompts import ChatPromptTemplate
from langchain.vectorstores import Chroma
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnableLambda, RunnablePassthrough
import os
import gradio as gr
import whisper
from openai import OpenAI

# *****Deploying the ChatBot using Gradio Application*****

# Defining the Function for Text-to-Text Conversational ChatBot
def conversation_chatbot_function(api_key, audio_path): #question):

  # Setting-Up OpenAI API Key
  OPENAI_API_KEY = api_key
  os.environ['OPENAI_API_KEY'] = OPENAI_API_KEY

  # Loading the CSV File
  loader = CSVLoader("conversation_data.csv", encoding="windows-1252")
  documents = loader.load()

  # Loading the OpenAIEmbeddings Model
  embedding_function = OpenAIEmbeddings()

  # Generating Embeddings (Vectors) for the Data present in the CSV File
  db = Chroma.from_documents(documents, embedding_function)
  retriever = db.as_retriever()

  # Defining the Prompt Template
  template = """Answer the question based only on the following context:
  {context}

  Question: {question}

  Note : Your reply should not be like 3rd person. You are the member of HAMSI Marketing and you should answer accordingly.

  Also if you do not know the answer reply with our Mail ID HAMSIMarketing@gmail.com and Contact Number 9999988888  stating to Contact us.
  """
  prompt = ChatPromptTemplate.from_template(template)

  # Loading the OpenAI LLM Model
  model = ChatOpenAI()

  # Defining the QA_Chain
  qa_chain = (
      {"context": retriever, "question": RunnablePassthrough()}
      | prompt
      | model
      | StrOutputParser()
  )

  # Speech-to-Text
  model_whisper = whisper.load_model("base")
  result = model_whisper.transcribe(audio_path)
  text_result = qa_chain.invoke(result["text"])

  # Loading the OPENAI Model
  client = OpenAI()

  # Text-to-Speech
  response = client.audio.speech.create(
      model="tts-1",
      voice="alloy",
      input=text_result,
  )

  # print(response)

  audio_result = response.stream_to_file("output.mp3")

  # return audio_result
  return "output.mp3"

# Launching Gradio App for Testing Speech to Text to Speech Function
audio_input = gr.components.Audio(type="filepath")


# Launching the Gradio Application

demo = gr.Interface(fn=conversation_chatbot_function, inputs=["textbox",audio_input], outputs="audio", title='Conversational ChatBot')

if __name__ == "__main__":
  demo.launch(debug=True)#inline=False)