Update app.py

app.py

@@ -1,25 +1,35 @@
 import os
-
-# Install required libraries
-os.system("pip install streamlit edge_tts pydub soxr numpy onnxruntime sentencepiece huggingface_hub torch beautifulsoup4 requests urllib3")
+# Install necessary libraries
+os.system('pip install streamlit torch onnxruntime transformers sentencepiece pydub soxr edge-tts requests beautifulsoup4')
 
 import streamlit as st
-import edge_tts
-import asyncio
-import tempfile
-import numpy as np
-import soxr
-from pydub import AudioSegment
 import torch
 import sentencepiece as spm
 import onnxruntime as ort
-from huggingface_hub import hf_hub_download, InferenceClient
+from pydub import AudioSegment
+import numpy as np
+import soxr
+import edge_tts
 import requests
 from bs4 import BeautifulSoup
 import urllib
 import random
+from huggingface_hub import hf_hub_download, InferenceClient
+import tempfile
+
+# Install necessary libraries
+os.system('pip install streamlit torch onnxruntime transformers sentencepiece pydub soxr edge-tts requests beautifulsoup4')
+
+# Load models
+model_name = "neongeckocom/stt_en_citrinet_512_gamma_0_25"
+sample_rate = 16000
+preprocessor = torch.jit.load(hf_hub_download(model_name, "preprocessor.ts", subfolder="onnx"))
+encoder = ort.InferenceSession(hf_hub_download(model_name, "model.onnx", subfolder="onnx"))
+tokenizer = spm.SentencePieceProcessor(hf_hub_download(model_name, "tokenizer.spm", subfolder="onnx"))
+
+client1 = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
+system_instructions1 = "<s>[SYSTEM] Answer as Real OpenGPT 4o, Made by 'KingNish', 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. 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]"
 
-# List of user agents to choose from for requests
 _useragent_list = [
     'Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:66.0) Gecko/20100101 Firefox/66.0',
     'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/111.0.0.0 Safari/537.36',
@@ -31,86 +41,58 @@ _useragent_list = [
 ]
 
 def get_useragent():
-    """Returns a random user agent from the list."""
     return random.choice(_useragent_list)
 
 def extract_text_from_webpage(html_content):
-    """Extracts visible text from HTML content using BeautifulSoup."""
     soup = BeautifulSoup(html_content, "html.parser")
-    # Remove unwanted tags
     for tag in soup(["script", "style", "header", "footer", "nav"]):
         tag.extract()
-    # Get the remaining visible text
     visible_text = soup.get_text(strip=True)
     return visible_text
 
-def search(term, num_results=1, lang="en", advanced=True, sleep_interval=0, timeout=5, safe="active", ssl_verify=None):
-    """Performs a Google search and returns the results."""
+def search(term, num_results=1):
     escaped_term = urllib.parse.quote_plus(term)
     start = 0
     all_results = []
 
-    # Fetch results in batches
     while start < num_results:
         resp = requests.get(
             url="https://www.google.com/search",
-            headers={"User-Agent": get_useragent()}, # Set random user agent
+            headers={"User-Agent": get_useragent()},
             params={
                 "q": term,
-                "num": num_results - start, # Number of results to fetch in this batch
-                "hl": lang,
+                "num": num_results - start,
+                "hl": "en",
                 "start": start,
-                "safe": safe,
+                "safe": "active",
             },
-            timeout=timeout,
-            verify=ssl_verify,
+            timeout=5,
         )
-        resp.raise_for_status() # Raise an exception if request fails
-
+        resp.raise_for_status()
         soup = BeautifulSoup(resp.text, "html.parser")
         result_block = soup.find_all("div", attrs={"class": "g"})
-
-        # If no results, continue to the next batch
         if not result_block:
             start += 1
             continue
 
-        # Extract link and text from each result
         for result in result_block:
             link = result.find("a", href=True)
             if link:
                 link = link["href"]
                 try:
-                    # Fetch webpage content
                     webpage = requests.get(link, headers={"User-Agent": get_useragent()})
                     webpage.raise_for_status()
-                    # Extract visible text from webpage
                     visible_text = extract_text_from_webpage(webpage.text)
                     all_results.append({"link": link, "text": visible_text})
                 except requests.exceptions.RequestException as e:
-                    # Handle errors fetching or processing webpage
-                    print(f"Error fetching or processing {link}: {e}")
                     all_results.append({"link": link, "text": None})
             else:
                 all_results.append({"link": None, "text": None})
 
-        start += len(result_block) # Update starting index for next batch
+        start += len(result_block)
 
     return all_results
 
-# Speech Recognition Model Configuration
-model_name = "neongeckocom/stt_en_citrinet_512_gamma_0_25"
-sample_rate = 16000
-
-# Download preprocessor, encoder and tokenizer
-preprocessor = torch.jit.load(hf_hub_download(model_name, "preprocessor.ts", subfolder="onnx"))
-encoder = ort.InferenceSession(hf_hub_download(model_name, "model.onnx", subfolder="onnx"))
-tokenizer = spm.SentencePieceProcessor(hf_hub_download(model_name, "tokenizer.spm", subfolder="onnx"))
-
-# Mistral Model Configuration
-client1 = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
-system_instructions1 = "<s>[SYSTEM] Answer as Real OpenGPT 4o, Made by 'KingNish', 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. 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]"
-
 def resample(audio_fp32, sr):
     return soxr.resample(audio_fp32, sr, sample_rate)
 
@@ -128,7 +110,7 @@ def transcribe(audio_path):
     input_signal = torch.tensor(audio_16k).unsqueeze(0)
     length = torch.tensor(len(audio_16k)).unsqueeze(0)
     processed_signal, _ = preprocessor.forward(input_signal=input_signal, length=length)
-    
+
     logits = encoder.run(None, {'audio_signal': processed_signal.numpy(), 'length': length.numpy()})[0][0]
 
     blank_id = tokenizer.vocab_size()
@@ -138,8 +120,7 @@ def transcribe(audio_path):
     return text
 
 def model(text, web_search):
-    if web_search is True:
-        """Performs a web search, feeds the results to a language model, and returns the answer."""
+    if web_search:
         web_results = search(text)
         web2 = ' '.join([f"Link: {res['link']}\nText: {res['text']}\n\n" for res in web_results])
         formatted_prompt = system_instructions1 + text + "[WEB]" + str(web2) + "[OpenGPT 4o]"
@@ -162,11 +143,33 @@ async def respond(audio, web_search):
 # Streamlit interface
 st.title("OpenGPT 4o DEMO")
 
-web_search = st.checkbox("Web Search", value=False)
-input_audio = st.file_uploader("Upload Audio", type=["wav", "mp3", "ogg"])
-if st.button("Transcribe and Respond"):
-    if input_audio:
-        output_audio_path = asyncio.run(respond(input_audio, web_search))
-        audio_file = open(output_audio_path, "rb")
-        audio_bytes = audio_file.read()
-        st.audio(audio_bytes, format="audio/wav")
+# Chat input interface
+st.subheader("💬 SuperChat")
+prompt = st.text_input("Say something")
+if prompt:
+    web_search = st.checkbox("Web Search", value=True)
+    response = model(prompt, web_search)
+    st.write(response)
+
+# Audio input interface
+st.subheader("🗣️ Voice Chat")
+audio_file = st.file_uploader("Upload an audio file", type=["wav", "mp3"])
+if audio_file:
+    web_search = st.checkbox("Web Search", value=False)
+    with st.spinner("Transcribing and generating response..."):
+        audio_path = audio_file.name
+        with open(audio_path, "wb") as f:
+            f.write(audio_file.getbuffer())
+        response_audio = await respond(audio_path, web_search)
+        st.audio(response_audio)
+    ```
+
+### Explanation:
+1. **Library Installation**: Uses `os.system()` to install the required libraries at the beginning of the script.
+2. **Streamlit Application**: The rest of the script remains the same as previously explained, including the Streamlit UI and the functionalities.
+
+### How to Run:
+Save this script as `app.py` and run it using the following command:
+
+```bash
+streamlit run app.py