Update app.py

app.py

@@ -2,104 +2,54 @@ import streamlit as st
 from audio_recorder_streamlit import audio_recorder
 import speech_recognition as sr
 from gtts import gTTS
-import os
 import tempfile
-from langchain.chains import RetrievalQA
-from langchain.llms import Anthropic  # Claude's API from Langchain
-from langchain.vectorstores import FAISS
-from langchain.embeddings.openai import OpenAIEmbeddings
-
-# Claude (Anthropic) API setup for Langchain
-CLAUDE_API_KEY = st.secrets['claude']['api_key']
-llm = Anthropic(model="claude-v1", api_key=CLAUDE_API_KEY)
+import os
+from anthropic import Client  # For Claude Haiku model
 
-# Setup the vector store for RAG
-embedding_model = OpenAIEmbeddings()
-vector_store = FAISS.load_local("my_vector_store", embedding_model)
+# Claude API setup
+CLAUDE_API_KEY = st.secrets['claude_api_key']  # Store your Claude API key in Streamlit secrets
+client = Client(api_key=CLAUDE_API_KEY)
 
-# Main function
+# Main function for chatbot app
 def main():
-    st.markdown(
-        """
-        <style>
-        .stApp {
-            background: linear-gradient(90deg, rgba(36,0,2,1) 0%, rgba(121,9,9,1) 50%, rgba(255,212,0,1) 100%);
-        }
-        .header-text {
-            color: white;
-            text-shadow: 2px 2px 4px #000000;
-            text-align: center;
-        }
-        .audio-recorder-container {
-            padding: 20px;
-            border: 2px solid #ccc;
-            border-radius: 10px;
-            background-color: #f9f9f9;
-            display: flex;
-            justify-content: center;
-            align-items: center;
-            position: relative;
-        }
-        </style>
-        """, 
-        unsafe_allow_html=True
-    )
-
-    st.markdown("<h1 class='header-text'>🎤 اردو وائس چیٹ بوٹ</h1>", unsafe_allow_html=True)
-    st.subheader('اپنی آواز ریکارڈ کریں اور اردو جواب سنیں', divider='rainbow')
-
-    st.sidebar.image('Hamesh_Raj_Profile_Photo.png', use_column_width=True)
-    st.sidebar.header("**انجینئر ہمیش راج**")
-    st.sidebar.write("جنریٹیو اے آئی اور مشین لرننگ میں دو سال کا تجربہ")
-
-    st.sidebar.header("چیٹ بوٹ کی معلومات", divider='rainbow')
-    st.sidebar.write('یہ چیٹ بوٹ اردو زبان میں صوتی پیغامات وصول کرتا ہے اور صوتی پیغامات میں ہی جواب دیتا ہے۔')
-
-    st.sidebar.header("رابطہ کی معلومات", divider='rainbow')
-    st.sidebar.write("[LinkedIn](https://www.linkedin.com/in/datascientisthameshraj/)")
-    st.sidebar.write("[GitHub](https://github.com/mldatascientist23)")
-    st.sidebar.write("[Email]([email protected])")
-
-    with st.container():
-        st.markdown('<div class="audio-recorder-container">', unsafe_allow_html=True)
-    # Audio recorder for Urdu input
-    audio_data = audio_recorder(text='بولیۓ', icon_size="2x", icon_name="microphone-lines", key="urdu_recorder")
-    st.markdown('</div>', unsafe_allow_html=True)
-
+    st.title("🎤 اردو وائس چیٹ بوٹ")
+    
+    # Sidebar with information
+    st.sidebar.title("حامش راج")
+    st.sidebar.write("ماہر ڈیٹا سائنس اور جنریٹو اے آئی")
+    
+    st.markdown("**اپنی آواز ریکارڈ کریں اور جواب حاصل کریں**")
+    
+    # Audio Recorder
+    audio_data = audio_recorder(text='اردو میں بولیئے', icon_size="2x", icon_name="microphone-lines", key="urdu_recorder")
+    
     if audio_data is not None:
-        with st.container():
-            col1, col2 = st.columns(2)
+        # Save the recorded audio to a temporary file
+        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_audio_file:
+            temp_audio_file.write(audio_data)
+            temp_audio_file_path = temp_audio_file.name
 
-            with col2:
-                # Display the recorded audio file
-                st.markdown('<h2 class="avatar">🧑</h2>', unsafe_allow_html=True)
-                st.audio(audio_data)
-                
-                # Save the recorded audio to a temporary file
-                with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_audio_file:
-                    temp_audio_file.write(audio_data)
-                    temp_audio_file_path = temp_audio_file.name
+        # Convert audio to text (Speech to Text in Urdu)
+        user_input_text = convert_audio_to_text(temp_audio_file_path)
 
-                # Convert audio file to text
-                text = convert_audio_to_text(temp_audio_file_path)
-                st.markdown(f'<div class="user-text">{text}</div>', unsafe_allow_html=True)
+        # Display user input text
+        st.write(f"**آپ نے کہا:** {user_input_text}")
+        
+        # Get LLM (Claude) response
+        response_text = get_claude_response(user_input_text)
 
-                # Remove the temporary file
-                os.remove(temp_audio_file_path)
+        # Display chatbot's text response
+        st.write(f"**جواب:** {response_text}")
 
-        # Get response from the Langchain RAG model
-        response_text = get_llm_response(text)
+        # Convert response text to audio and play it
+        response_audio = convert_text_to_audio(response_text)
+        st.audio(response_audio)
 
-        with st.container():
-            col1, col2 = st.columns(2)
-
-            with col1:
-                # Convert the response text to speech
-                convert_text_to_audio(response_text)
-                st.markdown(f'<div class="bot-text">{response_text}</div>', unsafe_allow_html=True)
+        # Clean up temporary audio file
+        os.remove(temp_audio_file_path)
 
+# Function to convert audio to text (Urdu Speech Recognition)
 def convert_audio_to_text(audio_file_path):
-    # Convert Urdu audio to text using speech recognition
     recognizer = sr.Recognizer()
     with sr.AudioFile(audio_file_path) as source:
         audio_data = recognizer.record(source)
@@ -107,26 +57,33 @@ def convert_audio_to_text(audio_file_path):
             text = recognizer.recognize_google(audio_data, language="ur")
             return text
         except sr.UnknownValueError:
-            return "آپ کی آواز واضح نہیں ہے"
+            return "معذرت، میں آپ کی آواز سمجھ نہیں سکا"
         except sr.RequestError:
-            return "معذرت، میری سروس دستیاب نہیں ہے"
+            return "معذرت، سرور دستیاب نہیں ہے"
 
-def convert_text_to_audio(text, lang='ur'):
+# Function to get response from Claude (Langchain with RAG)
+def get_claude_response(prompt_text):
+    prompt = f"براہ کرم اردو میں جواب دیں: {prompt_text}"
     try:
-        tts = gTTS(text=text, lang=lang)
-        tts_audio_path = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False).name
-        tts.save(tts_audio_path)
-        st.audio(tts_audio_path, format='audio/mp3')
+        # Query Claude Haiku via Langchain
+        response = client.completions.create(
+            model="claude-v1",  # Claude Haiku model
+            prompt=prompt,
+            max_tokens_to_sample=100,
+        )
+        return response['completion']
     except Exception as e:
-        st.error(f"آواز میں تبدیل کرنے میں خرابی: {e}")
+        return f"خطا: {str(e)}"
 
-def get_llm_response(text):
-    # Create the retrieval-augmented generation chain
-    qa_chain = RetrievalQA(llm=llm, retriever=vector_store.as_retriever())
-    
-    # Generate the response
-    response = qa_chain.run(text)
-    return response
+# Function to convert text to Urdu audio (Text-to-Speech)
+def convert_text_to_audio(text):
+    try:
+        tts = gTTS(text=text, lang='ur')
+        temp_audio_path = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False).name
+        tts.save(temp_audio_path)
+        return temp_audio_path
+    except Exception as e:
+        return None
 
 if __name__ == "__main__":
     main()