Spaces:
Sleeping
Sleeping
File size: 2,262 Bytes
0441ef7 ef1d0d3 0441ef7 2895ffe 0441ef7 2895ffe 0441ef7 1918a73 0441ef7 1918a73 0441ef7 ef1d0d3 71de052 0441ef7 bf88de2 71de052 0441ef7 71de052 0441ef7 |
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 |
import spaces
import tempfile
import gradio as gr
from streaming_stt_nemo import Model
from huggingface_hub import InferenceClient
import edge_tts
default_lang = "en"
engines = {default_lang: Model(default_lang)}
def transcribe(audio):
lang = "en"
model = engines[lang]
text = model.stt_file(audio)[0]
return text
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
system_instructions = "[SYSTEM] You are CrucialCoach, an AI-powered conversational coach. Guide the user through challenging workplace situations using the principles from 'Crucial Conversations'. Ask one question at a time and provide step-by-step guidance.\n\n[USER]"
@spaces.GPU(duration=120)
def model(text):
generate_kwargs = dict(
temperature=0.7,
max_new_tokens=512,
top_p=0.95,
repetition_penalty=1,
do_sample=True,
seed=42,
)
formatted_prompt = system_instructions + text + "[CrucialCoach]"
stream = client.text_generation(
formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
if not response.token.text == "</s>":
output += response.token.text
return output
async def respond(audio):
user = transcribe(audio)
reply = model(user)
communicate = edge_tts.Communicate(reply)
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
tmp_path = tmp_file.name
await communicate.save(tmp_path)
return tmp_path
theme = gr.themes.Base()
with gr.Blocks() as voice:
with gr.Row():
input = gr.Audio(label="Voice Chat", sources="microphone", type="filepath", waveform_options=False)
output = gr.Audio(label="CrucialCoach", type="filepath",
interactive=False,
autoplay=True,
elem_classes="audio")
gr.Interface(
fn=respond,
inputs=[input],
outputs=[output], live=True)
with gr.Blocks(theme=theme, css="footer {visibility: hidden}textbox{resize:none}", title="CrucialCoach DEMO") as demo:
gr.TabbedInterface([voice], ['🗣️ Crucial Coach Chat'])
demo.queue(max_size=200)
demo.launch() |