import gradio as gr
import torch
import librosa
from transformers import Wav2Vec2ForSequenceClassification, Wav2Vec2FeatureExtractor

model_name = "greenarcade/wav2vec2-vd-bird-sound-classification"
# model = Wav2Vec2ForSequenceClassification.from_pretrained(
#     model_name,
#     local_files_only=False,
#     use_auth_token=None,
#     trust_remote_code=False
# )
model = Wav2Vec2ForSequenceClassification.from_pretrained(model_name)
feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained(model_name)

def predict(audio_file):
    # Handle MP3/WAV files
    audio, sr = librosa.load(audio_file, sr=16000)

    # Process audio
    inputs = feature_extractor(
        audio,
        sampling_rate=16000,
        return_tensors="pt",
        padding=True,
        truncation=True,
        max_length=16000 * 5,
    )

    # Predict
    with torch.no_grad():
        logits = model(**inputs).logits
    probs = torch.softmax(logits, dim=-1).squeeze().tolist()

    # Format results - return actual float values instead of formatted strings
    predictions = {model.config.id2label[i]: prob for i, prob in enumerate(probs)}
    sorted_preds = sorted(predictions.items(), key=lambda x: x[1], reverse=True)[:3]
    return {k: v for k, v in sorted_preds}


# Gradio Interface
demo = gr.Interface(
    fn=predict,
    inputs=gr.Audio(sources=["upload"], type="filepath"),
    outputs=gr.Label(num_top_classes=3),
    title="🦜 Bird Sound Classifier (Indian birds)",
    description="Upload a 5-second audio clip to identify bird species",
    examples=[["greyheron-sample.wav"], ["blue-tail-sample.mp3"]]
)

demo.launch()