streaming voice

app.py

@@ -7,7 +7,7 @@ from huggingface_hub import hf_hub_download
 from llama_cpp import Llama
 from faster_whisper import WhisperModel
 
-from utilsinference import get_sentence, tts_interface, generate_llm_output
+from utilsasync import get_sentence, tts_interface
 
 os.environ["CUDACXX"] = "/usr/local/cuda/bin/nvcc"
 os.system('python -m unidic download')
@@ -61,16 +61,14 @@ def respond(chat_history, voice):
     if not voice:
         return None, gr.Warning("Please select a voice.")
     
-    sentence = generate_llm_output(chat_history[-1][0], chat_history[:-1], mistral_llm)
-    audiopb = tts_interface(sentence, voice)
-
-    print("Inserting sentence to queue")
-    print(sentence)
-
+    for sentence, chatbot_history in get_sentence(chat_history, mistral_llm):
+        print("Inserting sentence to queue")
+        print(sentence)
 
+        audiopb = tts_interface(sentence, voice)
     
     #history, response = get_sentence(chat_history, mistral_llm)
-    yield chat_history, sentence, audiopb
+        yield chatbot_history, sentence, audiopb
 
 
 #Gradio Interface