app.py

import json
import os
import tempfile

import gradio as gr
import gradio_client.utils as gc_utils
import pandas as pd

from src import batch

_orig_json_schema_to_python_type = gc_utils._json_schema_to_python_type


def _safe_json_schema_to_python_type(schema, defs):
    if isinstance(schema, bool):
        return "Any"
    return _orig_json_schema_to_python_type(schema, defs)


gc_utils._json_schema_to_python_type = _safe_json_schema_to_python_type

RESULT_COLUMNS = [
    "name",
    "emotional_tone",
    "emotional_intensity",
    "background_noise_present",
    "background_noise_type",
    "background_noise_severity",
    "audio_quality",
    "speaker_overlap_present",
    "long_silence_present",
    "confidence",
]


def run_batch(file, progress=gr.Progress()):
    if file is None:
        return pd.DataFrame(columns=RESULT_COLUMNS), pd.DataFrame(columns=["file", "error"]), None, None

    def on_progress(done, total, name):
        progress(done / max(total, 1), desc=f"Analyzing {name} ({done}/{total})")

    results_df, errors = batch.process_batch(file.name, on_progress=on_progress)
    if results_df.empty:
        results_df = pd.DataFrame(columns=RESULT_COLUMNS)
    errors_df = pd.DataFrame(errors) if errors else pd.DataFrame(columns=["file", "error"])

    out_dir = tempfile.mkdtemp()
    csv_path = os.path.join(out_dir, "results.csv")
    json_path = os.path.join(out_dir, "results.json")
    results_df.to_csv(csv_path, index=False)
    with open(json_path, "w") as f:
        json.dump(results_df.to_dict(orient="records"), f, indent=2)

    return results_df, errors_df, csv_path, json_path


with gr.Blocks(title="AutoAce Voice Tone & Background Noise") as demo:
    gr.Markdown(
        "# AutoAce Voice Tone & Background Noise Dashboard\n"
        "Upload a `.zip` containing the audio files and a `labels.csv` / manifest "
        "at the root (manifest is optional for unlabeled batches). "
        "Supported audio: wav, mp3, ogg, flac, m4a."
    )
    upload = gr.File(label="Evaluation batch (.zip)", file_types=[".zip"])
    run_btn = gr.Button("Run analysis", variant="primary")

    results_table = gr.Dataframe(label="Results", headers=RESULT_COLUMNS, wrap=True)
    errors_table = gr.Dataframe(label="Errors / skipped files", headers=["file", "error"], wrap=True)

    with gr.Row():
        csv_out = gr.File(label="Download results.csv")
        json_out = gr.File(label="Download results.json")

    run_btn.click(run_batch, inputs=[upload], outputs=[results_table, errors_table, csv_out, json_out])


if __name__ == "__main__":
    user = os.environ.get("AUTOACE_USER", "autoace")
    password = os.environ.get("AUTOACE_PASSWORD")
    auth = (user, password) if password else None
    demo.queue().launch(auth=auth, server_name="0.0.0.0", show_api=False)