Create app.py

app.py

@@ -0,0 +1,183 @@
+from langchain.chains import RetrievalQA
+from langchain_qdrant import QdrantVectorStore
+from langchain_community.embeddings import OpenAIEmbeddings
+from langchain_community.chat_models import ChatOpenAI
+from langchain.prompts import PromptTemplate
+
+# Import additional libraries for web application and audio processing
+import streamlit as st
+import gradio as gr
+import speech_recognition as sr
+from gtts import gTTS
+from io import BytesIO
+import base64
+from streamlit_chat import message
+
+import os
+os.environ['OPENAI_API_KEY'] = os.environ.get("OPENAI_API_KEY") 
+Qdrant_API_KEY = os.environ.get("Qdrant_API_KEY")
+
+
+# Initialize embedding model for vector representation of text
+embed_model = OpenAIEmbeddings()
+
+# Define the Qdrant URL for vector storage
+qdrant_url = "https://ee4d124d-d295-4df3-ad6b-47fe60d3f80d.europe-west3-0.gcp.cloud.qdrant.io:6333"
+
+# Create a Qdrant vector store from the document chunks and embeddings
+# vectorstore = QdrantVectorStore.from_documents(chunks, embed_model, url=qdrant_url, api_key=Qdrant_API_KEY, collection_name="Pakistan_Constitution_eng")
+vectorstore = QdrantVectorStore.from_existing_collection(embedding=embed_model, url=qdrant_url, api_key=Qdrant_API_KEY, collection_name="Pakistan_Constitution_eng") # read from existing collection
+# Initialize the language model for chat interactions
+llm = ChatOpenAI(model_name="gpt-4o-mini", temperature=0)
+
+# Build the prompt template for the QA system
+template = """You are a legal assistant providing accurate and concise information based on the Constitution of Pakistan. 
+You will receive questions in Urdu language and you have to answer in Urdu language. 
+Use the following pieces of context to answer the question at the end. 
+Give accurate references to the articles and clauses. 
+If you don't know the answer, just say that you don't know, don't try to make up an answer. 
+Provide proper relevant citations or references to the exact articles or clauses in the Constitution. 
+Keep the answer as concise as possible. Always say "thanks!" in urdu at the end of the answer.
+{context}
+Question: {question}
+Helpful Answer:"""
+
+# Define the prompt template for the retrieval QA chain
+QA_CHAIN_PROMPT = PromptTemplate(input_variables=["context", "question"], template=template)
+
+# Create a RetrievalQA chain using the language model and vector store
+qa_chain = RetrievalQA.from_chain_type(
+    llm,
+    retriever=vectorstore.as_retriever(),
+    chain_type_kwargs={"prompt": QA_CHAIN_PROMPT}
+)
+
+# Function to convert text to speech and return as base64-encoded audio
+def speak_urdu(text):
+    tts = gTTS(text=text, lang='ur')  # Convert text to speech in Urdu
+    audio_fp = BytesIO()  # Create a file-like object in memory
+    tts.write_to_fp(audio_fp)  # Write the audio to the object
+    audio_fp.seek(0)  # Reset pointer to the beginning
+    audio_bytes = audio_fp.read()  # Read the audio bytes
+    audio_base64 = base64.b64encode(audio_bytes).decode()  # Encode bytes to base64
+    return audio_base64
+
+# Function to recognize Urdu speech
+def recognize_speech():
+    recognizer = sr.Recognizer()  # Initialize the speech recognizer
+    with sr.Microphone() as source:
+        st.write("Listening...")  # Indicate that the app is listening
+        audio = recognizer.listen(source)  # Listen for audio input
+        try:
+            # Recognize speech using Google's API in Urdu
+            recognized_text = recognizer.recognize_google(audio, language="ur")
+            return recognized_text
+        except sr.UnknownValueError:
+            return "Sorry, I couldn't understand your speech."  # Handle unrecognized speech
+        except sr.RequestError:
+            return "Error: Unable to process your request."  # Handle request error
+
+# Placeholder function for Gradio chatbot interaction
+def invoke_chatbot(user_input):
+    response = qa_chain.invoke(user_input)  # Invoke the QA chain with user input
+    return response["result"]  # Return the result from the chain
+
+# Helper function to autoplay audio using HTML
+def autoplay_audio(audio_base64):
+    audio_html = f"""
+        <audio autoplay="true" controls="false" style="display:none;">
+        <source src="data:audio/mp3;base64,{audio_base64}" type="audio/mp3">
+        </audio>
+    """
+    st.markdown(audio_html, unsafe_allow_html=True)  # Render audio in Streamlit
+
+# Initialize session state variables for chat history and input
+if "history" not in st.session_state:
+    st.session_state["history"] = []  # Chat history
+
+if "input_text" not in st.session_state:
+    st.session_state["input_text"] = ""  # User input text
+
+if "voice_input" not in st.session_state:
+    st.session_state["voice_input"] = ""  # Voice input text
+
+if "audio_playback" not in st.session_state:
+    st.session_state["audio_playback"] = None  # Current audio for playback
+
+# Clear chat function to reset session state
+def clear_chat():
+    st.session_state["history"] = []  # Clear chat history
+    st.session_state["input_text"] = ""  # Clear user input text
+    st.session_state["audio_playback"] = None  # Clear audio playback
+
+# Sidebar for developer details, disclaimer, and copyright information
+st.sidebar.image("c.png", use_column_width=True)
+st.sidebar.title("Developer Details")
+st.sidebar.write("Developed by: **Abdul S.**")
+st.sidebar.write("@XevenSolutions")
+st.sidebar.write("LinkedIn: [linkedin.com/in/kushikhlaq](https://www.linkedin.com/in/kushikhlaq/)")
+
+st.sidebar.title("Disclaimer")
+st.sidebar.write("This chatbot provides information on the Constitution of Pakistan in Urdu. "
+                 "Please note that the information may not be comprehensive or up to date. "
+                 "For official references, please consult legal professionals.")
+
+st.sidebar.title("Copyright")
+st.sidebar.write("© 2024 Abdul Samad. All rights reserved.")
+
+# Streamlit app layout for user interaction
+st.title("Urdu Chatbot - Constitution of Pakistan")
+st.write("Chat with the Constitution of Pakistan in Urdu, either by typing or speaking.")
+
+# Button to clear chat
+if st.button("Clear Chat"):
+    clear_chat()  # Call the clear chat function
+
+# Handle user text input
+def handle_user_input():
+    user_input = st.session_state["input_text"]  # Get user input from session state
+    if user_input:
+        # Pass the user input to the chatbot model
+        chatbot_response = invoke_chatbot(user_input)
+
+        # Store the chat history
+        st.session_state["history"].append({"user": user_input, "bot": chatbot_response})
+
+        # Generate audio for the chatbot response
+        audio_base64 = speak_urdu(chatbot_response)
+        st.session_state["audio_playback"] = audio_base64  # Store the audio for playback
+
+        # Clear the input after submission
+        st.session_state["input_text"] = ""
+
+# Text input field with callback to handle user input
+st.text_input(
+    "Type your question in Urdu",
+    value=st.session_state["input_text"],  # Default value from session state
+    key="input_text",  # Key for session state
+    on_change=handle_user_input  # Callback function on input change
+)
+
+# Voice input option to recognize speech and handle it
+if st.button("Speak Now"):
+    recognized_text = recognize_speech()  # Get recognized speech
+    if recognized_text:
+        # Handle voice input by appending it to chat history
+        chatbot_response = invoke_chatbot(recognized_text)
+        st.session_state["history"].append({"user": recognized_text, "bot": chatbot_response})
+
+        # Generate audio for the chatbot response
+        audio_base64 = speak_urdu(chatbot_response)
+        st.session_state["audio_playback"] = audio_base64  # Store the audio for playback
+
+        st.session_state["voice_input"] = ""  # Clear voice input after processing
+
+# Display chat history using streamlit_chat component
+for idx, chat in enumerate(st.session_state["history"]):
+    message(chat["user"], is_user=True, key=f"user_{idx}")  # Display user message
+    message(chat["bot"], key=f"bot_{idx}")  # Display bot response
+
+# Autoplay audio for the latest chatbot response
+if st.session_state["audio_playback"]:
+    autoplay_audio(st.session_state["audio_playback"])  # Play the audio
+    st.session_state["audio_playback"] = None  # Reset audio playback state