Update utils_hf.py

utils_hf.py

@@ -1,14 +1,26 @@
+# import numpy as np
+# import soundfile as sf
+# import torchaudio
+# from transformers import pipeline
+
 import os
 import yt_dlp
 import streamlit as st
 import torch
 import whisper
-import requests
+import google.generativeai as genai
 from pyannote.audio import Pipeline
 from pyannote.audio.pipelines.utils.hook import ProgressHook
 from moviepy.editor import AudioFileClip
+from openai import Client
+from dotenv import load_dotenv
+
+load_dotenv()
 
-hf_token = os.getenv("HF_DIARIZATION_TOKEN")
+hf_token = os.environ.get("HF_DIARIZATION_TOKEN")
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+st.info(f"Usando dispositivo: {device}")
 
 pyannote_pipeline = Pipeline.from_pretrained(
     "pyannote/speaker-diarization-3.1",
@@ -66,8 +78,9 @@ def process_transcripts(diarization_text, speakers):
 
 
 def summarize_speaker_transcripts(speaker_transcripts):
-    url = "https://api.gemini.com/v1/generateContent"
-    api_key = "AIzaSyDHZZoMRUe7-VZgIZtCVWsRznkoQys1hrE"
+    gapi_key = os.environ.get("GEMINI_API_KEY")
+    genai.configure(api_key=gapi_key)
+    model = genai.GenerativeModel("gemini-1.5-flash")
 
     for speaker, transcripts in speaker_transcripts.items():
         full_text = ' '.join(transcripts)
@@ -75,24 +88,11 @@ def summarize_speaker_transcripts(speaker_transcripts):
         if len(full_text) > max_tokens:
             full_text = full_text[:max_tokens]
 
-        data = {
-            "prompt": f"Por favor, resume el siguiente texto:\n\n{full_text}",
-            "model": "gemini-1.5-pro",
-            "max_tokens": 150
-        }
+        prompt=f"Por favor, resume el siguiente texto:\n\n{full_text}"
 
-        headers = {
-            "Authorization": f"Bearer {api_key}",
-            "Content-Type": "application/json"
-        }
+        response = model.generate_content(prompt)
 
-        response = requests.post(url, json=data, headers=headers)
-
-        if response.status_code == 200:
-            result = response.json()
-            return result["text"]
-        else:
-            st.error(f"{response.status_code}, {response.text}")
+        st.text_area(f"Resumen del Hablante {speaker}", response.text, height=200)
 
 
 def diarize_full_audio(audio_path):
@@ -147,7 +147,7 @@ def transcribe_audio_whisper_lib(audio_path):
 
 
 def download_youtube_audio(url, progress_bar):
-
+    # No incluyas la extensión '.wav' en el 'outtmpl'
     output_path = os.path.join('temp_audio')
     if os.path.exists(output_path + '.wav'):
         os.remove(output_path + '.wav')
@@ -223,3 +223,4 @@ class StreamlitProgressHook(ProgressHook):
         else:
             progress_message = f"{step_name:<20} ━ Progress data unavailable"
         self.progress_text.text(progress_message)
+