File size: 3,791 Bytes
22621c3 cf197d6 65948ef 22621c3 3b43e04 22621c3 3b43e04 cf197d6 3b43e04 5590000 3b43e04 5590000 3b43e04 5590000 cf197d6 22621c3 cf197d6 22621c3 3b43e04 22621c3 65948ef 22621c3 cf197d6 22621c3 cf197d6 3b43e04 65948ef 3b43e04 cf197d6 2b08886 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 |
import streamlit as st
from audio_recorder_streamlit import audio_recorder
from groq import Groq
import os
from langchain_groq import ChatGroq
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
import edge_tts
import asyncio
from dotenv import load_dotenv
load_dotenv()
#Front end using streamlit
def frontend():
st.title("Voice AI Demo")
# Initialize session state variables
if "conversation" not in st.session_state:
st.session_state.conversation = [] # Stores (question, answer, audio_filename)
if "audio_count" not in st.session_state:
st.session_state.audio_count = 1 # Start numbering audio files from output1.wav
status_placeholder = st.empty()
status_placeholder.write("Press Mic button to start asking a question")
recorded_audio = audio_recorder(sample_rate=8000)
text = st.chat_input()
def process_input(user_input):
status_placeholder.write("Getting response...")
response = answer(user_input)
status_placeholder.write("Converting response to audio...")
# Generate unique audio filename
audio_filename = f"output{st.session_state.audio_count}.wav"
asyncio.run(convert_audio(response, audio_filename))
st.session_state.audio_count += 1 # Increment for next response
status_placeholder.write("Press mic button again to ask more questions")
# Append (question, answer, audio_filename) to conversation history
st.session_state.conversation.append((f"Q: {user_input}", f"A: {response}", audio_filename))
# Handle user input
if text:
process_input(text)
elif recorded_audio:
status_placeholder.write("Converting audio...")
data_to_file(recorded_audio)
status_placeholder.write("Uploading audio...")
transcription = audio_to_text("temp_audio.wav")
status_placeholder.write("Transcription completed.")
process_input(transcription)
# Display full conversation history
for i, (q, a, audio_file) in enumerate(st.session_state.conversation):
st.write(q)
st.write(a)
st.audio(audio_file, format="audio/wav", loop=False, autoplay=(i == len(st.session_state.conversation) - 1))
#Fuction to convert audio data to audio file
def data_to_file(recorded_audio):
temp_audio_path = "temp_audio.wav"
with open(temp_audio_path, "wb") as temp_file:
temp_file.write(recorded_audio)
#Function for audio to text
def audio_to_text(audio_path):
client = Groq(api_key=os.getenv('GROQ_API_KEY'))
with open(audio_path, 'rb') as file:
transcription = client.audio.translations.create(
file=(audio_path, file.read()),
model='whisper-large-v3',
)
return transcription.text
#Function for answerig User Query
def answer(user_question):
model = ChatGroq(
model="llama-3.3-70b-versatile",
temperature=0.6
)
prompt = ChatPromptTemplate([
("system", "You are super knowlegable AI chat bot which will answer all User Query, answer with confident, also this response will get convert back to speech, so dont make point or anything, but make your answer in para form and dont make it too large, and use proper annotation, comma, full stop, question mark, so that a better text to speach can be genrate back."),
("user", "User Query: {question}"),
])
parser = StrOutputParser()
chain = prompt|model|parser
answer = chain.invoke({'question': user_question})
return answer
# Audio conversion
async def convert_audio(text, filename):
voice = "fr-FR-VivienneMultilingualNeural"
communicate = edge_tts.Communicate(text, voice)
await communicate.save(filename)
frontend() |