app.py

import gradio as gr
from transformers import pipeline, set_seed
from gtts import gTTS
import uuid
import os
import speech_recognition as sr

# Inisialisasi model Bahasa Indonesia
generator = pipeline("text-generation", model="flax-community/gpt2-small-indonesian")
set_seed(42)

# Fungsi chatbot: teks masuk → jawaban + TTS
def chat_with_bot(prompt):
    result = generator(prompt, max_length=100, num_return_sequences=1)[0]["generated_text"]
    response = result[len(prompt):].strip()
    
    # Buat audio dari jawaban
    audio_path = f"/tmp/{uuid.uuid4().hex}.mp3"
    tts = gTTS(response, lang="id")
    tts.save(audio_path)
    
    return response, audio_path

# Fungsi untuk input teks atau suara
def transcribe_and_respond(text_input, audio_input):
    if audio_input:
        recognizer = sr.Recognizer()
        with sr.AudioFile(audio_input) as source:
            audio_data = recognizer.record(source)
        try:
            text_input = recognizer.recognize_google(audio_data, language="id-ID")
        except:
            return "⚠️ Gagal mengenali suara.", None

    if not text_input.strip():
        return "⚠️ Tidak ada input yang diproses.", None

    return chat_with_bot(text_input)

# Gradio UI dengan tampilan mobile-friendly
with gr.Blocks(css="""
.gradio-container {
    max-width: 600px !important;
    margin: auto;
    padding: 16px;
    font-family: 'Segoe UI', sans-serif;
}
h1, h2 {
    text-align: center;
}
textarea, input {
    font-size: 16px !important;
}
@media screen and (max-width: 600px) {
    .gradio-container {
        padding: 12px;
    }
}
""") as demo:
    gr.Markdown("""
    <h1>MEDVA – Medical Virtual Assistant</h1>
    <p style='text-align:center;'>Tulis pertanyaan atau unggah suara Anda. MEDVA akan menjawab dan membacakan hasilnya!</p>
    """)

    with gr.Column():
        text_input = gr.Textbox(label="💬 Masukkan Teks (opsional)", placeholder="Contoh: Apa itu AI?", lines=2)
        audio_input = gr.Audio(label="🎤 Upload Suara (opsional)", type="filepath")
        submit_btn = gr.Button("🚀 Kirim")

    with gr.Column():
        text_output = gr.Textbox(label="🧠 Jawaban", lines=3)
        audio_output = gr.Audio(label="🔊 Jawaban Audio")

    submit_btn.click(fn=transcribe_and_respond, inputs=[text_input, audio_input], outputs=[text_output, audio_output])

demo.launch()