interview_copilot / app copy 8.py
alex buz
new
68ba2e8
raw
history blame
4.59 kB
import time
import gradio as gr
from transformers import pipeline
import numpy as np
from openai import OpenAI
import threading
import queue
transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-base.en")
qa_model = pipeline("question-answering", model="distilbert-base-cased-distilled-squad")
class PubSub:
def __init__(self):
self.subscribers = []
def subscribe(self, callback):
self.subscribers.append(callback)
def publish(self, message):
for subscriber in self.subscribers:
subscriber(message)
def predict(message, history, api_key, is_paused, pubsub):
def run_prediction():
client = OpenAI(api_key=api_key)
history_openai_format = []
for human, assistant in history:
history_openai_format.append({"role": "user", "content": human})
history_openai_format.append({"role": "assistant", "content": assistant})
history_openai_format.append({"role": "user", "content": message})
response = client.chat.completions.create(
model='gpt-4o',
messages=history_openai_format,
temperature=1.0,
stream=True
)
partial_message = ""
for chunk in response:
if is_paused[0]:
while is_paused[0]:
time.sleep(0.1)
if chunk.choices[0].delta.content:
partial_message += chunk.choices[0].delta.content
pubsub.publish(partial_message)
thread = threading.Thread(target=run_prediction)
thread.start()
def chat_with_api_key(api_key, message, history, is_paused):
pubsub = PubSub()
result_queue = queue.Queue()
def update_message(partial_message):
result_queue.put(partial_message)
pubsub.subscribe(update_message)
predict(message, history, api_key, is_paused, pubsub)
while True:
try:
accumulated_message = result_queue.get(timeout=0.1)
history.append((message, accumulated_message))
yield message, [[message, accumulated_message]]
except queue.Empty:
if not any(thread.is_alive() for thread in threading.enumerate() if thread != threading.current_thread()):
break
def transcribe(audio):
if audio is None:
return "No audio recorded."
sr, y = audio
y = y.astype(np.float32)
y /= np.max(np.abs(y))
return transcriber({"sampling_rate": sr, "raw": y})["text"]
def answer(transcription):
context = "You are a chatbot answering general questions"
result = qa_model(question=transcription, context=context)
return result['answer']
def process_audio(audio):
if audio is None:
return "No audio recorded.", []
transcription = transcribe(audio)
answer_result = answer(transcription)
return transcription, [[transcription, answer_result]]
def update_output(api_key, audio_input, state, is_paused):
if is_paused[0]:
yield "", state
else:
message = transcribe(audio_input)
responses = chat_with_api_key(api_key, message, state, is_paused)
for response, updated_state in responses:
if is_paused[0]:
break
yield response, updated_state
def clear_all():
return None, "", []
def toggle_pause(is_paused):
is_paused[0] = not is_paused[0]
return is_paused
def update_button_label(is_paused):
return "Resume" if is_paused[0] else "Pause"
with gr.Blocks() as demo:
answer_output = gr.Chatbot(label="Answer Result")
with gr.Row():
audio_input = gr.Audio(label="Audio Input", sources=["microphone"], type="numpy")
with gr.Column():
api_key = gr.Textbox(label="API Key", placeholder="Enter your API key", type="password")
transcription_output = gr.Textbox(label="Transcription")
clear_button = gr.Button("Clear")
pause_button = gr.Button("Pause")
state = gr.State([])
is_paused = gr.State([False])
audio_input.stop_recording(
fn=update_output,
inputs=[api_key, audio_input, state, is_paused],
outputs=[transcription_output, answer_output]
)
clear_button.click(
fn=clear_all,
inputs=[],
outputs=[audio_input, transcription_output, answer_output]
)
pause_button.click(
fn=toggle_pause,
inputs=[is_paused],
outputs=[is_paused]
).then(
fn=update_button_label,
inputs=[is_paused],
outputs=[pause_button]
)
demo.launch()