add coqui xtts option

app.py

@@ -20,7 +20,10 @@ with gr.Blocks() as main_app:
                 chat = gr.Chatbot(label="Chatbot Crunchyroll")
                 output_video = gr.Video(interactive=False, label='Video', autoplay=True, height=400)
 
-        output_audio = gr.Audio(interactive=False, label='Audio', autoplay=False)
+        with gr.Column():
+            with gr.Row():
+                options_audio = gr.Radio(["XTTS", "Elevenlabs"], label="Audio Generation")
+                output_audio = gr.Audio(interactive=False, label='Audio', autoplay=False)
 
         messages = gr.State([])
 
@@ -35,7 +38,7 @@ with gr.Blocks() as main_app:
     with gr.Tab('Times'):
 
         columns = ["User Message", "Chatbot Response", "Standalone Question", "Create Embedding", "Query Pinecone",
-                   "Context Prompt", "Final Response GPT", "Create Clean Message", "Create Audio Elevenlabs", "Create Video D-iD", "Final Time"]
+                   "Context Prompt", "Final Response GPT", "Create Clean Message", "Create Audio", "Create Video", "Final Time"]
         table_times = gr.DataFrame(headers=columns, visible=False, interactive=False)
 
         with gr.Column():
@@ -46,7 +49,7 @@ with gr.Blocks() as main_app:
 
     text.submit(
         fn=get_answer,
-        inputs=[text, chat, messages, output_audio, output_video, table_times],
+        inputs=[text, chat, messages, output_audio, output_video, table_times, options_audio],
         outputs=[chat, output_audio, output_video, table_times]
     ).then(
         lambda: None, None, [text]
@@ -58,7 +61,7 @@ with gr.Blocks() as main_app:
 
     button_text.click(
         fn=get_answer,
-        inputs=[text, chat, messages, output_audio, output_video, table_times],
+        inputs=[text, chat, messages, output_audio, output_video, table_times, options_audio],
         outputs=[chat, output_audio, output_video, table_times]
     ).then(
         lambda: None, None, [text]

backend_functions.py

@@ -15,6 +15,8 @@ from pymongo.mongo_client import MongoClient
 from utils import create_folders
 from gcp import download_credentials
 from csv import writer
+import asyncio
+import httpx
 from dotenv import load_dotenv
 load_dotenv()
 
@@ -38,6 +40,7 @@ IMG_XAVY = os.getenv("IMG_XAVY")
 CREDENTIALS_GCP = os.getenv("GOOGLE_APPLICATION_CREDENTIALS")
 NAME_BUCKET = os.getenv("NAME_BUCKET")
 
+URL_AUDIO = os.getenv("URL_AUDIO")
 
 # Chat
 openai_client = OpenAI(api_key=OPENAI_API_KEY)
@@ -155,6 +158,8 @@ def _get_standalone_question(question, history_messages):
             print(e)
     
     prompt_standalone = file_prompt_standalone.replace('HISTORY', history).replace('QUESTION', question)
+    print(prompt_standalone)
+    print("------------------")
     standalone_msg_q = _call_gpt_standalone(prompt_standalone)
     print(standalone_msg_q)
     print("------------------")
@@ -167,47 +172,66 @@ def _create_clean_message(text: str):
     return clean_answer
 
 
-def _create_audio(clean_text: str):
+async def _create_audio(clean_text: str, option_audio: str):
     download_credentials()
     create_folders()
     
     STORAGE_CLIENT = storage.Client.from_service_account_json(CREDENTIALS_GCP)
 
     unique_id = str(uuid.uuid4())
+    signed_url_audio = "None"
 
-    # Create audio file
-    client_elevenlabs = ElevenLabs(api_key=API_KEY_ELEVENLABS)
-    voice_custom = Voice(voice_id = "ZQe5CZNOzWyzPSCn5a3c")
+    if option_audio == "Elevenlabs":
+        # Create audio file with elevenlabs
+        client_elevenlabs = ElevenLabs(api_key=API_KEY_ELEVENLABS)
+        voice_custom = Voice(voice_id = "ZQe5CZNOzWyzPSCn5a3c")
 
-    audio = client_elevenlabs.generate(
-        text=clean_text,
-        voice=voice_custom,
-        model="eleven_multilingual_v2"
-    )
+        audio = client_elevenlabs.generate(
+            text=clean_text,
+            voice=voice_custom,
+            model="eleven_multilingual_v2"
+        )
 
-    source_audio_file_name = f'./audios/file_audio_{unique_id}.wav'
+        source_audio_file_name = f'./audios/file_audio_{unique_id}.wav'
 
-    try:
-        save(audio, source_audio_file_name)
-    except Exception as e:
-        print(e)
+        try:
+            save(audio, source_audio_file_name)
+        except Exception as e:
+            print(e)
 
-    # Save audio and get url of gcp
-    destination_blob_name_audio = unique_id + '.wav'
-    
-    bucket = STORAGE_CLIENT.bucket(NAME_BUCKET)
-    blob = bucket.blob(destination_blob_name_audio)
-    try:
-        blob.upload_from_filename(source_audio_file_name)
-    except Exception as e:
-        print(e)
+        # Save audio and get url of gcp
+        destination_blob_name_audio = unique_id + '.wav'
+        
+        bucket = STORAGE_CLIENT.bucket(NAME_BUCKET)
+        blob = bucket.blob(destination_blob_name_audio)
+        try:
+            blob.upload_from_filename(source_audio_file_name)
+        except Exception as e:
+            print(e)
 
-    signed_url_audio = "None"
-    try:
-        url_expiration = timedelta(minutes=15)
-        signed_url_audio = blob.generate_signed_url(expiration=url_expiration)
-    except Exception as e:
-        print(e)
+        try:
+            url_expiration = timedelta(minutes=15)
+            signed_url_audio = blob.generate_signed_url(expiration=url_expiration)
+        except Exception as e:
+            print(e)
+    
+    elif option_audio == "XTTS":
+        params = {'text': clean_text, 'language': 'es'}
+        headers = {'accept': 'application/json'}
+
+        # Makes a request to the instance with the audio api
+        async with httpx.AsyncClient() as client:
+            try:
+                response = await client.get(URL_AUDIO, params=params, headers=headers, timeout=120)
+            except Exception as e:
+                print(f'There is a problem with the audio. Check that instance. ERROR: {e}')
+
+        # Check if everything was successful
+        if response.status_code == 200:
+            r = response.json()
+            signed_url_audio = r['link_audio']
+        else:
+            print(f'There is a problem with the audio. Check that instance. ERROR: {response.status_code}')
 
     return signed_url_audio, unique_id
 
@@ -296,7 +320,7 @@ def _create_video(link_audio: str, unique_id: str):
     return signed_url_video
 
 
-def get_answer(question: str, chatbot: list[tuple[str, str]], history_messages, comp_audio, comp_video, df_table):
+def get_answer(question: str, chatbot: list[tuple[str, str]], history_messages, comp_audio, comp_video, df_table, option_audio):
     """
     Gets the answer of the chatbot
     """
@@ -338,7 +362,7 @@ def get_answer(question: str, chatbot: list[tuple[str, str]], history_messages,
     time_create_clean_message = end_create_clean_message - start_create_clean_message
     
     start_create_audio = time.time()
-    url_audio, unique_id = _create_audio(processed_message) # create audio with elevenlabs
+    url_audio, unique_id = asyncio.run(_create_audio(processed_message, option_audio)) # create audio
     end_create_audio = time.time()
     time_create_audio = end_create_audio - start_create_audio
 
@@ -372,7 +396,7 @@ def get_answer(question: str, chatbot: list[tuple[str, str]], history_messages,
 
 def init_greeting(chatbot, history_messages):
     if len(chatbot) == 0:
-        greeting = ('Hola 👋, soy Roll, tu asistente de recomendación de series y películas animadas en Crunchyroll. ¿En qué puedo ayudarte hoy?')
+        greeting = ('Hola 👋, soy tu asistente de recomendación de series y películas animadas en Crunchyroll. ¿En qué puedo ayudarte hoy?')
         history_messages.append({'role': 'assistant', 'content': greeting})
         chatbot.append([None, greeting])