api.py

import torch, io, torchaudio
import numpy as np
from pipeline import BioacousticEngine
engine = None

def get_engine():
    global engine
    if engine is None: engine = BioacousticEngine()
    return engine

def predict_bird(audio_bytes, distance_threshold=0.8):
    try:
        ae = get_engine()
        waveform, sample_rate = torchaudio.load(io.BytesIO(audio_bytes))
        processed_wave = ae.process_waveform(waveform, sample_rate)
        with torch.no_grad():
            mel_spec = ae.preprocessor.process(processed_wave).unsqueeze(0)
            if mel_spec.shape[-1] >= 184: mel_spec = mel_spec[:, :, :, :184]
            else: mel_spec = torch.nn.functional.pad(mel_spec, (0, 184 - mel_spec.shape[-1]))
            embedding = ae.model(mel_spec).cpu().numpy().flatten().reshape(1, -1)
        coords = ae.reducer.transform(embedding)
        x, y = coords[0][0], coords[0][1]
        valid_df = ae.df[ae.df['Cluster_ID'] != -1].copy()
        distances = np.sqrt((valid_df['UMAP_X'] - x)**2 + (valid_df['UMAP_Y'] - y)**2)
        min_dist = distances.min()
        if min_dist > distance_threshold:
            return {"status": "NO_BIRD_DETECTED", "x": float(x), "y": float(y), "distance": float(min_dist)}
        match_row = valid_df.loc[distances.idxmin()]
        confidence = max(0.0, (distance_threshold - min_dist) / distance_threshold) * 100
        return {"status": "SUCCESS", "species": match_row['Target_Bird'].upper(), "confidence": round(confidence, 2), "x": float(x), "y": float(y), "anchor_id": match_row['File_ID']}
    except Exception as e:
        return {"status": "ERROR", "message": str(e)}