Update app.py

app.py

@@ -19,29 +19,28 @@ hub = {
     "HF_API_TOKEN": os.environ.get("HUGGINGFACE_API_TOKEN")
 }
 
-# Global state to keep memory
+# Global state
 vector_db = None
 qa_chain = None
-chat_memory = None  # Stores previous conversations
+chat_memory = None
 
-# Function to transcribe and initialize RAG pipeline
+# Transcribe and set up RAG
 def transcribe_and_setup(audio_file_path):
     global vector_db, qa_chain, chat_memory
 
     if audio_file_path is None:
-        return "No audio uploaded.", None
+        return "No audio uploaded.", []
 
-    # Transcribe with Whisper
     result = model.transcribe(audio_file_path)
     transcript = result["text"]
 
-    # Split and embed transcript
+    # Split and embed
     text_splitter = RecursiveCharacterTextSplitter(chunk_size=1024, chunk_overlap=64)
     splits = text_splitter.create_documents([transcript])
     embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
     vector_db = FAISS.from_documents(splits, embeddings)
 
-    # Create retriever + LLM QA chain with memory
+    # QA setup
     chat_memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
     retriever = vector_db.as_retriever()
     llm = HuggingFaceEndpoint(
@@ -53,24 +52,39 @@ def transcribe_and_setup(audio_file_path):
     )
     qa_chain = ConversationalRetrievalChain.from_llm(llm=llm, retriever=retriever, memory=chat_memory)
 
-    return "Transcription complete! Ready for questions.", None  # Don't return transcript
+    return "Transcription complete! Ready for questions.", []  # Empty chat reset
 
-# Function to handle chat with memory
+# Handle conversation
 def answer_question(question, chat_history):
     global qa_chain, chat_memory
     if qa_chain is None:
         return "Please upload and process an audio file first.", chat_history
 
-    response = qa_chain.invoke({"question": question, "chat_history": chat_memory.load_memory_variables({})["chat_history"]})
-    
-    # Append new interaction to chat history
-    chat_history.append(("User", question))
-    chat_history.append(("Assistant", response["answer"]))
-
-    return "", chat_history  # Return empty input box, updated chat history
+    response = qa_chain.invoke({
+        "question": question,
+        "chat_history": chat_memory.load_memory_variables({})["chat_history"]
+    })
+
+    # Just show back-and-forth messages
+    chat_history.append([question, response["answer"]])
+    return "", chat_history
+
+# Custom CSS
+custom_css = """
+.chatbox .message.user, .chatbox .message.bot {
+    background-color: #1e3d2f !important;
+    color: #ffffff !important;
+    border-radius: 10px !important;
+    padding: 10px !important;
+    margin: 5px !important;
+}
+.chatbox .message.user::before, .chatbox .message.bot::before {
+    content: none !important;
+}
+"""
 
-# Gradio UI
-with gr.Blocks(theme=gr.themes.Soft(), css="footer {display:none !important;}") as demo:
+# Gradio app
+with gr.Blocks(theme=gr.themes.Soft(), css=custom_css) as demo:
     gr.Markdown("## 🎙️ **Conversational Audio Chatbot**")
     gr.Markdown("Upload an audio file, let the AI process it, and ask any questions!")
 
@@ -80,20 +94,20 @@ with gr.Blocks(theme=gr.themes.Soft(), css="footer {display:none !important;}")
             transcribe_button = gr.Button("🚀 Process Audio")
             status_output = gr.Textbox(label="🛠️ Status", interactive=False)
         with gr.Column(scale=2):
-            chatbot = gr.Chatbot(label="💬 Chat with your audio", height=400)
+            chatbot = gr.Chatbot(label="💬 Chat with your audio", elem_classes=["chatbox"])
             question_input = gr.Textbox(label="Type your question", placeholder="Ask about the audio...")
             ask_button = gr.Button("💡 Ask")
 
     transcribe_button.click(
         fn=transcribe_and_setup,
         inputs=audio_input,
-        outputs=[status_output, chatbot]  # No transcript, just chatbot reset
+        outputs=[status_output, chatbot]
     )
 
     ask_button.click(
         fn=answer_question,
         inputs=[question_input, chatbot],
-        outputs=[question_input, chatbot]  # Keeps chat history alive
+        outputs=[question_input, chatbot]
     )
 
-demo.launch()
+demo.launch()