app.py

import gradio as gr
from app.pipeline import FullEgyptianVoiceChatbot

bot = FullEgyptianVoiceChatbot()

def text_chat(text):
    r = bot.chat_text(text)
    return r["user_text"], r["user_normalized"], r["assistant_text_raw"], r["assistant_text_tts"], r["audio_path"]

def voice_chat(audio_fp):
    r = bot.chat_voice(audio_fp)
    if "error" in r:
        return "", "", r["error"], "", None
    return r["user_text"], r["user_normalized"], r["assistant_text_raw"], r["assistant_text_tts"], r["audio_path"]

with gr.Blocks() as demo:
    gr.Markdown("Egyptian Arabic Voice Chatbot (Whisper ASR + JAIS + EGTTS)")

    with gr.Tab("Text"):
        inp = gr.Textbox(label="Input (Egyptian Arabic)")
        o_user = gr.Textbox(label="User Text")
        o_norm = gr.Textbox(label="Normalized")
        o_raw  = gr.Textbox(label="JAIS Raw Response")
        o_tts  = gr.Textbox(label="TTS-ready Response")
        o_aud  = gr.Audio(label="Bot Audio", type="filepath")
        btn = gr.Button("Send")
        btn.click(text_chat, inputs=inp, outputs=[o_user, o_norm, o_raw, o_tts, o_aud])

    with gr.Tab("Voice"):
        v_in = gr.Audio(sources=["microphone","upload"], type="filepath", label="Record / Upload")
        v_user = gr.Textbox(label="User Text (ASR)")
        v_norm = gr.Textbox(label="Normalized")
        v_raw  = gr.Textbox(label="JAIS Raw Response")
        v_tts  = gr.Textbox(label="TTS-ready Response")
        v_out  = gr.Audio(label="Bot Audio", type="filepath")
        v_btn = gr.Button("Send Voice")
        v_btn.click(voice_chat, inputs=v_in, outputs=[v_user, v_norm, v_raw, v_tts, v_out])

demo.queue()
demo.launch()