app.py

import gradio as gr
import os
from huggingface_hub import InferenceClient

HUGGINGFACE_API_KEY = os.environ["HUGGINGFACE_API_KEY"]


def transcript_audio(audio_file) -> str:
    model = "openai/whisper-large-v3"
    api = InferenceClient(model, token=HUGGINGFACE_API_KEY)
    text = api.automatic_speech_recognition(audio_file)
    return text


def summarize_text(text: str, bullet_points: int, conclusion: bool) -> str:
    llm_model = "google/gemma-7b-it"
    api = InferenceClient(llm_model, token=HUGGINGFACE_API_KEY)
    if conclusion:
        prompt = f"Summarize the following text into {bullet_points} bullet points and a conclusion: {text}"
    else:
        prompt = f"Summarize the following text into {bullet_points} bullet points: {text}"
    summary = api.text_generation(prompt, max_new_tokens=250, do_sample=True)
    return summary["generated_text"]

def control(audio_file: gr.AudioFile, text: str, bullet_points: int, conclusion: bool) -> str:
    if audio_file:
        text = transcript_audio(audio_file)
    summary = summarize_text(text, bullet_points, conclusion)
    return summary
# make a simeple interface, where the user can input a text and get a summary or input an audio file and get a transcript and a summary
iface = gr.Interface(
    fn=summarize_text,
    inputs=[
        gr.components.Audio(label="Audio file"),
        gr.components.Textbox(lines=5, label="Text"),
        gr.components.Slider(minimum=1, maximum=10, value=5, step=1, label="Number of bullet points"),
        gr.components.Checkbox(label="Add conclusion"),
    ],
    outputs=gr.components.Textbox(label="Summary")
)

iface.launch()