import os
import io
import librosa
import numpy as np
import logging
import soundfile
import torchaudio
import subprocess
from flask import Flask, request, send_file, render_template

logging.getLogger('numba').setLevel(logging.WARNING)
logging.getLogger('markdown_it').setLevel(logging.WARNING)
logging.getLogger('urllib3').setLevel(logging.WARNING)
logging.getLogger('matplotlib').setLevel(logging.WARNING)

template_dir = os.path.abspath('.')  # Путь к корневой папке
app = Flask(__name__, template_folder=template_dir)

limitation = os.getenv("SYSTEM") == "spaces"  # limit audio length in huggingface spaces

@app.route('/', methods=['GET'])
def index():
    speakers = ["chapaev", "petka", "anka", "narrator", "floppa"]
    return render_template('index.html', speakers=speakers, limitation=limitation)

@app.route('/inference', methods=['POST'])
def run_inference():
    input_audio = request.files['audio']
    speaker = request.form['speaker']

    if input_audio is None:
        return "You need to upload an audio", 400

    audio, sampling_rate = librosa.load(input_audio, sr=None)
    duration = audio.shape[0] / sampling_rate

    if duration > 120 and limitation:
        return "Please upload an audio file that is less than 120 seconds. If you need to generate a longer audio file, please use Colab.", 400

    if len(audio.shape) > 1:
        audio = librosa.to_mono(audio.transpose(1, 0))
    if sampling_rate != 16000:
        audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=16000)

    # TODO edit from GUI
    cluster_ratio = 1
    noise_scale = 2
    is_pitch_prediction_enabled = True
    f0_method = "dio"
    transpose = 0

    model_path = f"./models/{speaker}/{speaker}.pth"
    config_path = f"./models/{speaker}/config.json"
    cluster_path = ""

    raw_path = 'tmp.wav'
    soundfile.write(raw_path, audio, 16000, format="wav")

    inference_cmd = f"svc infer {raw_path} -m {model_path} -c {config_path} {f'-k {cluster_path} -r {cluster_ratio}' if cluster_path != '' and cluster_ratio > 0 else ''} -t {transpose} --f0-method {f0_method} -n {noise_scale} -o out.wav {'' if is_pitch_prediction_enabled else '--no-auto-predict-f0'}"
    print(inference_cmd)

    result = subprocess.run(
        inference_cmd.split(),
        stdout=subprocess.PIPE,
        stderr=subprocess.STDOUT,
        text=True
    )

    return send_file('out.wav', mimetype='audio/wav')

if __name__ == '__main__':
    app.run(host="0.0.0.0", port=7860)