app.py

import tempfile
from pathlib import Path

import openai
import streamlit as st
from streamlit_mic_recorder import mic_recorder
import os

openai.api_key = os.getenv('OPENAI_API_KEY')

st.set_page_config(page_title="", page_icon="🎙️", layout="wide")

st.markdown("<h1 style='text-align: center;'>🗣️ Покупатель Н. Должанский</h1>", unsafe_allow_html=True)
st.markdown("<p style='text-align: center;'>Показывает пользователю, как общаться</p>",
            unsafe_allow_html=True)

if "chat_history" not in st.session_state:
    st.session_state.chat_history = []


def read_file(path: str):
    with open(Path(path), 'r', encoding='utf-8') as f:
        return f.read()


prompt_1 = read_file('prompt_dolzhansky.txt')
prompt_2 = read_file('supervisor_prompt')


def transcribe_audio(audio_path):
    with open(audio_path, "rb") as audio:
        transcript = openai.audio.transcriptions.create(
            model="whisper-1",
            file=audio
        )
    return transcript.text


def chat_with_gpt(user_message, prompt):
    if not user_message:
        return ""
    hist = []
    for i in st.session_state.chat_history:
        hist.append(
            {'role': 'user', 'content': i['user']}
        )
        hist.append(
            {'role': 'assistant', 'content': i['assistant']}
        )
    hist.append({'role': 'user', 'content': user_message})
    response = openai.chat.completions.create(
        model="gpt-4o",
        messages=[{'role': 'system', 'content': prompt, }, ] + hist
    )

    gpt_reply = response.choices[0].message.content
    return gpt_reply



def generate_audio(text):
    tts_audio = openai.audio.speech.create(
        model="tts-1",
        voice="ash",
        input=text
    )

    temp_audio_path = tempfile.mktemp(suffix=".mp3")
    with open(temp_audio_path, "wb") as audio_file:
        audio_file.write(tts_audio.read())

    return temp_audio_path


with st.container():
    for entry in st.session_state.get("chat_history", []):
        with st.chat_message("user"):
            st.markdown(entry["user"])

        with st.chat_message("assistant"):
            st.markdown(entry["assistant"])
            try:
                st.audio(entry["audio"], format="audio/mp3")
            except:
                continue
                
inp = st.text_input("💬 Введите сообщение:", "")

user_input = None

audio_data = mic_recorder(start_prompt="🎤 Нажмите для записи", stop_prompt="🛑 Остановить")

if audio_data is not None:
    audio_path = tempfile.mktemp(suffix=".wav")
    with open(audio_path, "wb") as f:
        f.write(audio_data['bytes'])

    text_from_audio = transcribe_audio(audio_path)
    user_input = text_from_audio

if st.button("📨 Отправить"):
    if inp:
        response_text = chat_with_gpt(inp, prompt_1)
        audio_path = generate_audio(response_text)

        st.session_state.chat_history.append({
            "user": inp,
            "assistant": response_text,
            "audio": audio_path
        })

        st.rerun()
    if user_input:
        response_text = chat_with_gpt(user_input, prompt_1)
        audio_path = generate_audio(response_text)

        st.session_state.chat_history.append({
            "user": user_input,
            "assistant": response_text,
            "audio": audio_path
        })
        user_input = None
        audio_data = None
        st.rerun()
        
if st.button("📨 Получить оценку"):
        response_text = chat_with_gpt('Напиши оценку продавца.', prompt_2)

        st.session_state.chat_history.append({
            "user": user_input,
            "assistant": response_text,
        })
        user_input = None
        audio_data = None
        st.rerun()