app.py

from transformers import pipeline
from openai import OpenAI
import requests
from PIL import Image
from pydub import AudioSegment
import gradio as gr
import os
import json
import numpy as np



client = OpenAI(
    api_key=os.environ.get('OPEN_AI_KEY')
    #api_key="NULLas"
)

pipe = pipeline(model="Siphh/wablab2", revision="85f5ec9bd0813eec8b403a7fb510d067a4b49397")

psychologist_url = "https://raw.githubusercontent.com/rogoran/id2223kth.github.io/master/src/serverless-ml-intro/serverless-ml-lab/wine_images/360_F_169387942_tUJn9ADX31sRDFCuK9nlBuRR4gO6a1TK.jpg"
IMG = Image.open(requests.get(psychologist_url, stream=True).raw)
latest_messages = [{"role": "system", "content": "Du är en svensktalande psykolog."}]

def transcribe(audio):
    if audio is None or audio == "":
        print("NO AUDIO FOUND")
        return "", ""

    
    text_obj = pipe(audio)
    text = text_obj["text"]
    return chat(text), text
    
    
def chat(prompt):
    #response = generate_chat(prompt)
    latest_messages.append({"role": "user", "content": prompt})
    response = client.chat.completions.create(
        model="gpt-3.5-turbo-1106",
        messages=latest_messages
    )
    latest_messages.append({"role": "assistant", "content": response.choices[0].message.content})
    return response.choices[0].message.content  
    
with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            inp = gr.Audio(sources="microphone", type="filepath")
            btn = gr.Button("Send transcribed msg")
            transcr_text = gr.Textbox(label="Din senaste fråga")
        with gr.Column():
            gr.Image(value=IMG, label="Doctor Amanda")
            out = gr.Textbox(label="Doctor Amandas answers")
            btn.click(fn=transcribe, inputs=inp, outputs=[out, transcr_text])
            
demo.launch()