import os
import torch
import torchaudio
import tempfile
from transformers import WhisperProcessor, WhisperForConditionalGeneration
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import streamlit as st

os.environ["TRANSFORMERS_CACHE"] = "/app/cache"
os.makedirs("/app/cache", exist_ok=True)

whisper_processor = WhisperProcessor.from_pretrained("openai/whisper-tiny")
whisper_model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-tiny")
text_model = AutoModelForSequenceClassification.from_pretrained("bert-base-uncased")
tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")

def transcribe(audio_bytes):
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
        tmp.write(audio_bytes)
        tmp_path = tmp.name
    waveform, sample_rate = torchaudio.load(tmp_path)
    input_features = whisper_processor(waveform.squeeze().numpy(), sampling_rate=sample_rate, return_tensors="pt").input_features
    predicted_ids = whisper_model.generate(input_features)
    transcription = whisper_processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
    os.remove(tmp_path)
    return transcription

def extract_text_features(text):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
    outputs = text_model(**inputs)
    return outputs.logits.argmax(dim=1).item()

def predict