app.py

import gradio as gr
from alm_pipeline import analyze_audio
from reasoning import generate_reasoning

def full_pipeline(audio, question):
    if audio is None:
        return "No audio uploaded", ""

    summary = analyze_audio(audio)
    answer = generate_reasoning(summary, question)

    return summary, answer


ui = gr.Interface(
    fn=full_pipeline,
    inputs=[
        gr.Audio(sources=["upload"], type="filepath"),
        gr.Textbox(label="Ask a Question About the Audio")
    ],
    outputs=[
        gr.JSON(label="Audio Understanding Summary"),
        gr.Textbox(label="Reasoning Answer")
    ],
    title="ALM Prototype – Hugging Face Edition",
    description="Upload audio → detect speech, sounds → ask any question → get AI reasoning."
)

ui.launch()