Update app.py

app.py

@@ -1,98 +1,115 @@
-import os
-import re
 import gradio as gr
+from transformers import pipeline
+import requests
+import json
 import edge_tts
 import asyncio
-import time
 import tempfile
+import os
+import inflect
 from huggingface_hub import InferenceClient
+import re
+import time
+from streaming_stt_nemo import Model
 
-DESCRIPTION = """ # <center><b>JARVIS⚡</b></center>
-        ### <center>A personal Assistant of Tony Stark for YOU
-        ### <center>Currently It supports text input, But If this space completes 1k hearts than I starts working on Audio Input.</center>
-        """
-
-MORE = """ ## TRY Other Models
-        ### Instant Video: Create Amazing Videos in 5 Second -> https://huggingface.co/spaces/KingNish/Instant-Video
-        ### Instant Image: 4k images in 5 Second -> https://huggingface.co/spaces/KingNish/Instant-Image
-        """
+number_to_word = inflect.engine()
 
-Fast = """## Fastest Model"""
+default_lang = "en"
 
-Complex = """## Best in Complex Question"""
+engines = { default_lang: Model(default_lang) }
 
-Detail = """## Best for Detailed Generation or Long Answers"""
+def transcribe(audio):
+    lang = "en"
+    model = engines[lang]
+    text = model.stt_file(audio)[0]
+    return text
 
 client1 = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
 
 system_instructions1 = "<s>[SYSTEM] Answer as Real Jarvis JARVIS, Made by 'Tony Stark', Keep conversation very short, clear, friendly and concise. The text provided is a request for a specific type of response from you, the virtual assistant. The request asks you to provide friendly responses as if You are the character Jarvis, made by 'Tony Stark.' The expectation is that I will avoid introductions and start answering the query directly, elaborating on all aspects of the request. As an AI-powered assistant, your task is to generate responses that appear as if they are created by the user and not an AI-powered assistant. [USER]"
 
-async def generate1(prompt):
+def model(text):
     generate_kwargs = dict(
-        temperature=0.6,
-        max_new_tokens=256,
+        temperature=0.7,
+        max_new_tokens=512,
         top_p=0.95,
         repetition_penalty=1,
         do_sample=True,
         seed=42,
     )
-    formatted_prompt = system_instructions1 + prompt + "[JARVIS]"
+    
+    formatted_prompt = system_instructions1 + text + "[JARVIS]"
     stream = client1.text_generation(
-        formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
+        formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
     output = ""
     for response in stream:
         if not response.token.text == "</s>":
             output += response.token.text
 
-    communicate = edge_tts.Communicate(output)
-    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
-        tmp_path = tmp_file.name
-        await communicate.save(tmp_path)
-    yield tmp_path
+    return output
 
-client2 = InferenceClient("meta-llama/Meta-Llama-3-70B-Instruct")
+def number_to_words(str):
+    words = str.split(' ')
+    result = []
+    
+    for word in words:
+        if( any(char.isdigit() for char in word) ):
+            word = number_to_word.number_to_words(word)
 
-system_instructions2 = "[SYSTEM] Answer as Real Jarvis JARVIS, Made by 'Tony Stark', Must answer in friendly style and Easy Manner. You can answer Complex Questions. Do not say who are you or Hi, Hello, Just Start answering. Stop, as answer ends. [USER]"
+        result.append(word)
 
-async def generate2(prompt):
-    generate_kwargs = dict(
-        temperature=0.6,
-        max_new_tokens=512,
-        top_p=0.95,
-        repetition_penalty=1,
-        do_sample=True,
-    )    
-    formatted_prompt = system_instructions2 + prompt + "[ASSISTANT]"
-    stream = client2.text_generation(
-        formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
-    output = ""
-    for response in stream:
-        output += response.token.text
+    final_result = ' '.join(result).replace('point', '')
+    return final_result
 
-    communicate = edge_tts.Communicate(output)
+async def respond(audio):
+    user = transcribe(audio)
+    reply = model(user)
+    reply2 = number_to_words(reply)
+    communicate = edge_tts.Communicate(reply2)
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
         tmp_path = tmp_file.name
         await communicate.save(tmp_path)
     yield tmp_path
 
-client3 = InferenceClient("meta-llama/Meta-Llama-3-70B-Instruct")
+DESCRIPTION = """ # <center><b>JARVIS⚡</b></center>
+        ### <center>A personal Assistant of Tony Stark for YOU
+        ### <center>Voice Chat with your personal Assistant</center>
+        """
+
+MORE = """ ## TRY Other Models
+        ### Instant Video: Create Amazing Videos in 5 Second -> https://huggingface.co/spaces/KingNish/Instant-Video
+        ### Instant Image: 4k images in 5 Second -> https://huggingface.co/spaces/KingNish/Instant-Image
+        """
+
+BETA = """ ### Voice Chat (BETA)"""
+
+FAST = """## Fastest Model"""
+
+Complex = """## Best in Complex Question"""
+
+Detail = """## Best for Detailed Generation or Long Answers"""
+
+base_loaded = "mistralai/Mixtral-8x7B-Instruct-v0.1"
+
+client1 = InferenceClient(base_loaded)
 
-system_instructions3 = "[SYSTEM]The text provided is a request for a specific type of response from me, the virtual assistant. The request asks me to provide detailed and friendly responses as if I am the character Jarvis, inspired by 'Tony Stark.' The expectation is that I will avoid introductions and start answering the query directly, elaborating on all aspects of the request. As an AI-powered assistant, my task is to generate responses that appear as if they are created by the user and not an AI-powered assistant.[USER]"
+system_instructions1 = "[SYSTEM] Answer as Real Jarvis JARVIS, Made by 'Tony Stark', Keep conversation very short, clear, friendly and concise. The text provided is a request for a specific type of response from you, the virtual assistant. The request asks you to provide friendly responses as if You are the character Jarvis, made by 'Tony Stark.' The expectation is that I will avoid introductions and start answering the query directly, elaborating on all aspects of the request. As an AI-powered assistant, your task is to generate responses that appear as if they are created by the user and not an AI-powered assistant. [USER]"
 
-async def generate3(prompt):
+async def generate1(prompt):
     generate_kwargs = dict(
-        temperature=0.6,
-        max_new_tokens=2048,
+        temperature=0.7,
+        max_new_tokens=512,
         top_p=0.95,
         repetition_penalty=1,
-        do_sample=True,
-    )    
-    formatted_prompt = system_instructions3 + prompt + "[ASSISTANT]"
-    stream = client3.text_generation(
+        do_sample=False,
+    )
+    formatted_prompt = system_instructions1 + prompt + "[JARVIS]"
+    stream = client1.text_generation(
         formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
     output = ""
     for response in stream:
-        output += response.token.text
+        if not response.token.text == "</s>":
+            output += response.token.text
 
     communicate = edge_tts.Communicate(output)
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
@@ -102,6 +119,17 @@ async def generate3(prompt):
 
 with gr.Blocks(css="style.css") as demo:    
     gr.Markdown(DESCRIPTION)
+    with gr.Row():
+        user_input = gr.Audio(label="Voice Chat (BETA)".type="filepath")
+        output_audio = gr.Audio(label="JARVIS", type="filepath",
+                        interactive=False,
+                        autoplay=True,
+                        elem_classes="audio")
+    with gr.Row():
+        translate_btn = gr.Button("Response")
+        translate_btn.click(fn=respond, inputs=user_input,
+                            outputs=output_audio, api_name=False)
+    gr.Markdown(FAST)
     with gr.Row():
         user_input = gr.Textbox(label="Prompt", value="What is Wikipedia")
         input_text = gr.Textbox(label="Input Text", elem_id="important")
@@ -113,7 +141,7 @@ with gr.Blocks(css="style.css") as demo:
         translate_btn = gr.Button("Response")
         translate_btn.click(fn=generate1, inputs=user_input,
                             outputs=output_audio, api_name="translate")  
-        
+
 gr.Markdown(MORE)
 
 if __name__ == "__main__":