Update app.py

app.py

@@ -1,38 +1,86 @@
+from langchain_groq import ChatGroq
+import os
 import gradio as gr
 from langchain.chains import ConversationChain
 from langchain.memory import ConversationBufferMemory
-from langchain_groq import ChatGroq
+from langchain_community.embeddings.fastembed import FastEmbedEmbeddings
+from langchain.document_loaders import PyPDFLoader
+from langchain_experimental.text_splitter import SemanticChunker
+from langchain.vectorstores import FAISS
 from gtts import gTTS
 import tempfile
- 
-# Setup your LLM and memory
- 
+
+# Set your API key from Hugging Face Secrets
+# DO NOT hardcode your API key here
+GROQ_API_KEY = os.environ.get('GROQ_API_KEY')
+
+# Initialize Groq LLM
+llm = ChatGroq(
+    model_name="llama3-70b-8192",
+    temperature=0.7,
+    api_key=GROQ_API_KEY
+)
+
+# Initialize memory
+memory = ConversationBufferMemory()
 conversation = ConversationChain(llm=llm, memory=memory)
- 
-language_options = ['English', 'Hindi', 'Spanish', 'French', 'German', 'Tamil']
-menu_options = ["Places to Visit", "Best Time to Visit", "Festivals", "Cuisine", "Travel Tips"]
- 
-# Dummy vector search (replace with real embedding logic)
+
+# Load PDF and create embeddings
+def initialize_rag():
+    try:
+        # Load the PDF document
+        loader = PyPDFLoader("TourismChatbot.pdf")
+        pages = loader.load_and_split()
+        
+        # Create embeddings
+        embed_model = FastEmbedEmbeddings(model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
+        
+        # Create semantic chunks
+        semantic_chunker = SemanticChunker(embed_model, breakpoint_threshold_type="percentile")
+        semantic_chunks = semantic_chunker.create_documents([d.page_content for d in pages])
+        
+        # Create vector store
+        vectorstore = FAISS.from_documents(documents=semantic_chunks, embedding=embed_model)
+        
+        return vectorstore, embed_model
+    except Exception as e:
+        print(f"Error initializing RAG: {e}")
+        # Return None if initialization fails
+        return None, None
+
+# Initialize RAG components
+vectorstore, embed_model = initialize_rag()
+
+# Function to retrieve relevant information from the vector store
 def retrieve_relevant_chunks(query, top_k=3):
-    return ["Rajasthan is a state in India known for its forts, palaces, and desert landscapes."]
- 
+    try:
+        if vectorstore is not None:
+            documents = vectorstore.similarity_search(query, k=top_k)
+            return [doc.page_content for doc in documents]
+        else:
+            # Fallback content if vectorstore is not available
+            return ["Rajasthan is a state in India known for its forts, palaces, and desert landscapes."]
+    except Exception as e:
+        print(f"Error retrieving chunks: {e}")
+        return ["Rajasthan is a state in India known for its forts, palaces, and desert landscapes."]
+
 def generate_rag_response(query, language="English"):
     retrieved_chunks = retrieve_relevant_chunks(query)
     context = "\n".join(retrieved_chunks)
     prompt = f"""
     Please provide the answer in **{language}**.
- 
+
     You are a helpful AI assistant providing tourism information about Rajasthan.
     Answer based on the following context. If information is unavailable, say "I don't know."
- 
+
     Context: {context}
     Question: {query}
- 
+
     Answer:
     """
     response = conversation.run(prompt)
     return response.strip()
- 
+
 def generate_speech(text, language):
     lang_map = {"English": "en", "Hindi": "hi", "Spanish": "es", "French": "fr", "German": "de", "Tamil": "ta"}
     lang_code = lang_map.get(language, "en")
@@ -40,21 +88,26 @@ def generate_speech(text, language):
     temp_audio_path = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False).name
     tts.save(temp_audio_path)
     return temp_audio_path
- 
+
 def chatbot_interface(query, language, chat_history):
     response = generate_rag_response(query, language)
     speech_path = generate_speech(response, language)
- 
+
     # Just append a 🔊 icon and use Gradio to handle the file
     response_with_audio = f"{response} 🔊 (Click play below)"
     chat_history.append((query, response_with_audio))
- 
+
     return chat_history, speech_path, ""  # Return file path as separate gr.Audio
- 
+
 def handle_menu_click(topic, language, chat_history):
     query = f"Give me information about {topic} in Rajasthan."
     return chatbot_interface(query, language, chat_history)
- 
+
+# Define language and menu options
+language_options = ['English', 'Hindi', 'Spanish', 'French', 'German', 'Tamil']
+menu_options = ["Places to Visit", "Best Time to Visit", "Festivals", "Cuisine", "Travel Tips"]
+
+# Create the Gradio interface
 with gr.Blocks(css="""
     body {background-color: #FFF2E1; font-family: Arial, sans-serif;}
     .gradio-container {max-width: 800px; margin: auto; padding: 20px; background: #FFF2E1;
@@ -70,33 +123,28 @@ with gr.Blocks(css="""
         font-weight: bold; border-radius: 6px; padding: 5px 10px; cursor: pointer;}
     .chat-input {width: 100%; padding: 10px; border-radius: 8px; border: 1px solid #e1c7a6;}
 """) as demo:
- 
+
     gr.Markdown("<h2 class='gradio-title'>🐪 Rajasthan Tourism Chatbot</h2>")
- 
+
     language_selector = gr.Dropdown(language_options, value="English", label="Select Language")
     chatbot = gr.Chatbot(label="Rajasthan Tourism Assistant", elem_classes="gradio-chat")
- 
+
     with gr.Row():
         for topic in menu_options:
             btn = gr.Button(topic, elem_classes="gr-button")
             btn.click(handle_menu_click,
                       inputs=[gr.Textbox(value=topic, visible=False), language_selector, chatbot],
-                      outputs=[chatbot, gr.State(), gr.State()])
- 
+                      outputs=[chatbot, gr.Audio(label="🔊 Audio Response", type="filepath"), gr.Textbox()])
+
     query_input = gr.Textbox(placeholder="Ask about Rajasthan...", label="Enter your query", elem_classes="chat-input")
- 
-    def process_user_query(query, language, history):
-        updated_chat, _, _ = chatbot_interface(query, language, history)
-        return updated_chat, None, ""
- 
     audio_output = gr.Audio(label="🔊 Audio Response", type="filepath", visible=True)
- 
+
     query_input.submit(
         chatbot_interface,
         inputs=[query_input, language_selector, chatbot],
-        outputs=[chatbot, audio_output, gr.State()]
+        outputs=[chatbot, audio_output, query_input]
     )
- 
- 
-demo.launch(debug=True,share=True)
- 
+
+# Launch the app
+if __name__ == "__main__":
+    demo.launch(share=True)