Update app.py

app.py

@@ -2,7 +2,6 @@ import gradio as gr
 import os
 api_token = os.getenv("HF_TOKEN")
 
-
 from langchain_community.vectorstores import FAISS
 from langchain_community.document_loaders import PyPDFLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
@@ -18,11 +17,7 @@ import torch
 list_llm = ["meta-llama/Meta-Llama-3-8B-Instruct", "mistralai/Mistral-7B-Instruct-v0.2"]  
 list_llm_simple = [os.path.basename(llm) for llm in list_llm]
 
-# Load and split PDF document
 def load_doc(list_file_path):
-    # Processing for one document only
-    # loader = PyPDFLoader(file_path)
-    # pages = loader.load()
     loaders = [PyPDFLoader(x) for x in list_file_path]
     pages = []
     for loader in loaders:
@@ -34,14 +29,11 @@ def load_doc(list_file_path):
     doc_splits = text_splitter.split_documents(pages)
     return doc_splits
 
-# Create vector database
 def create_db(splits):
     embeddings = HuggingFaceEmbeddings()
     vectordb = FAISS.from_documents(splits, embeddings)
     return vectordb
 
-
-# Initialize langchain LLM chain
 def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, progress=gr.Progress()):
     if llm_model == "meta-llama/Meta-Llama-3-8B-Instruct":
         llm = HuggingFaceEndpoint(
@@ -77,36 +69,27 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, pr
     )
     return qa_chain
 
-# Initialize database
 def initialize_database(list_file_obj, progress=gr.Progress()):
-    # Create a list of documents (when valid)
     list_file_path = [x.name for x in list_file_obj if x is not None]
-    # Load document and create splits
     doc_splits = load_doc(list_file_path)
-    # Create or load vector database
     vector_db = create_db(doc_splits)
     return vector_db, "Database created!"
 
-# Initialize LLM
 def initialize_LLM(llm_option, llm_temperature, max_tokens, top_k, vector_db, progress=gr.Progress()):
-    # print("llm_option",llm_option)
     llm_name = list_llm[llm_option]
     print("llm_name: ",llm_name)
     qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, top_k, vector_db, progress)
     return qa_chain, "QA chain initialized. Chatbot is ready!"
 
-
 def format_chat_history(message, chat_history):
     formatted_chat_history = []
     for user_message, bot_message in chat_history:
         formatted_chat_history.append(f"User: {user_message}")
         formatted_chat_history.append(f"Assistant: {bot_message}")
     return formatted_chat_history
-    
 
 def conversation(qa_chain, message, history):
     formatted_chat_history = format_chat_history(message, history)
-    # Generate response using QA chain
     response = qa_chain.invoke({"question": message, "chat_history": formatted_chat_history})
     response_answer = response["answer"]
     if response_answer.find("Helpful Answer:") != -1:
@@ -115,14 +98,11 @@ def conversation(qa_chain, message, history):
     response_source1 = response_sources[0].page_content.strip()
     response_source2 = response_sources[1].page_content.strip()
     response_source3 = response_sources[2].page_content.strip()
-    # Langchain sources are zero-based
     response_source1_page = response_sources[0].metadata["page"] + 1
     response_source2_page = response_sources[1].metadata["page"] + 1
     response_source3_page = response_sources[2].metadata["page"] + 1
-    # Append user message and response to chat history
     new_history = history + [(message, response_answer)]
     return qa_chain, gr.update(value=""), new_history, response_source1, response_source1_page, response_source2, response_source2_page, response_source3, response_source3_page
-    
 
 def upload_file(file_obj):
     list_file_path = []
@@ -133,24 +113,23 @@ def upload_file(file_obj):
 
 def demo():
     custom_css = """
-    .horizontal-container {
-        display: flex !important;
-        flex-direction: row !important;
-        flex-wrap: nowrap !important;
-        width: 100% !important;
+    #column-container {
+        display: flex;
+        flex-direction: row;
+        flex-wrap: nowrap;
     }
-    .column-1 {
-        min-width: 300px !important;
-        max-width: 35% !important;
-        flex: 1 !important;
+    #column-left {
+        min-width: 350px;
+        max-width: 35%;
+        margin-right: 20px;
     }
-    .column-2 {
-        min-width: 500px !important;
-        flex: 2 !important;
+    #column-right {
+        min-width: 500px;
+        flex-grow: 1;
     }
-    @media (max-width: 900px) {
-        .column-1 { max-width: 40% !important; }
-        .column-2 { min-width: 400px !important; }
+    @media (max-width: 1200px) {
+        #column-left { min-width: 300px; }
+        #column-right { min-width: 400px; }
     }
     """
 
@@ -158,55 +137,41 @@ def demo():
         vector_db = gr.State()
         qa_chain = gr.State()
         gr.HTML("<center><h1>RAG PDF chatbot</h1><center>")
-        gr.Markdown("""<b>Query your PDF documents!</b> This AI agent is designed to perform retrieval augmented generation (RAG) on PDF documents. The app is hosted on Hugging Face Hub for the sole purpose of demonstration. \
-        <b>Please do not upload confidential documents.</b>
-        """)
+        gr.Markdown("""<b>Query your PDF documents!</b> This AI agent is designed to perform retrieval augmented generation (RAG) on PDF documents.""")
         
-        with gr.Row(elem_classes="horizontal-container"):
-            with gr.Column(elem_classes="column-1"):
+        with gr.Row(elem_id="column-container"):
+            with gr.Column(elem_id="column-left"):
                 gr.Markdown("<b>Step 1 - Upload PDF documents and Initialize RAG pipeline</b>")
-                with gr.Row():
-                    document = gr.Files(height=300, file_count="multiple", file_types=[".pdf"], interactive=True, label="Upload PDF documents")
-                with gr.Row():
-                    db_btn = gr.Button("Create vector database")
-                with gr.Row():
-                    db_progress = gr.Textbox(value="Not initialized", show_label=False)
-                gr.Markdown("<style>body { font-size: 16px; }</style><b>Select Large Language Model (LLM) and input parameters</b>")
-                with gr.Row():
-                    llm_btn = gr.Radio(list_llm_simple, label="Available LLMs", value=list_llm_simple[0], type="index")
-                with gr.Row():
-                    with gr.Accordion("LLM input parameters", open=False):
-                        with gr.Row():
-                            slider_temperature = gr.Slider(minimum=0.01, maximum=1.0, value=0.5, step=0.1, label="Temperature", info="Controls randomness in token generation", interactive=True)
-                        with gr.Row():
-                            slider_maxtokens = gr.Slider(minimum=128, maximum=9192, value=4096, step=128, label="Max New Tokens", info="Maximum number of tokens to be generated", interactive=True)
-                        with gr.Row():
-                            slider_topk = gr.Slider(minimum=1, maximum=10, value=3, step=1, label="top-k", info="Number of tokens to select the next token from", interactive=True)
-                with gr.Row():
-                    qachain_btn = gr.Button("Initialize Question Answering Chatbot")
-                with gr.Row():
-                    llm_progress = gr.Textbox(value="Not initialized", show_label=False)
-
-            with gr.Column(elem_classes="column-2"):
+                document = gr.Files(height=300, file_count="multiple", file_types=[".pdf"], interactive=True, label="Upload PDF documents")
+                db_btn = gr.Button("Create vector database")
+                db_progress = gr.Textbox(value="Not initialized", show_label=False)
+                gr.Markdown("<b>Select Large Language Model (LLM) and input parameters</b>")
+                llm_btn = gr.Radio(list_llm_simple, label="Available LLMs", value=list_llm_simple[0], type="index")
+                with gr.Accordion("LLM input parameters", open=False):
+                    slider_temperature = gr.Slider(0.01, 1.0, value=0.5, step=0.1, label="Temperature")
+                    slider_maxtokens = gr.Slider(128, 9192, value=4096, step=128, label="Max New Tokens")
+                    slider_topk = gr.Slider(1, 10, value=3, step=1, label="top-k")
+                qachain_btn = gr.Button("Initialize Question Answering Chatbot")
+                llm_progress = gr.Textbox(value="Not initialized", show_label=False)
+
+            with gr.Column(elem_id="column-right"):
                 gr.Markdown("<b>Step 2 - Chat with your Document</b>")
                 chatbot = gr.Chatbot(height=505)
                 with gr.Accordion("Relevant context from the source document", open=False):
                     with gr.Row():
-                        doc_source1 = gr.Textbox(label="Reference 1", lines=2, container=True, scale=20)
+                        doc_source1 = gr.Textbox(label="Reference 1", lines=2, scale=20)
                         source1_page = gr.Number(label="Page", scale=1)
                     with gr.Row():
-                        doc_source2 = gr.Textbox(label="Reference 2", lines=2, container=True, scale=20)
+                        doc_source2 = gr.Textbox(label="Reference 2", lines=2, scale=20)
                         source2_page = gr.Number(label="Page", scale=1)
                     with gr.Row():
-                        doc_source3 = gr.Textbox(label="Reference 3", lines=2, container=True, scale=20)
+                        doc_source3 = gr.Textbox(label="Reference 3", lines=2, scale=20)
                         source3_page = gr.Number(label="Page", scale=1)
-                with gr.Row():
-                    msg = gr.Textbox(placeholder="Ask a question", container=True)
+                msg = gr.Textbox(placeholder="Ask a question")
                 with gr.Row():
                     submit_btn = gr.Button("Submit")
                     clear_btn = gr.ClearButton([msg, chatbot], value="Clear")
-        
-        # Rest of your event handlers remain the same...
+
         db_btn.click(initialize_database,
                     inputs=[document],
                     outputs=[vector_db, db_progress])
@@ -216,7 +181,6 @@ def demo():
                         inputs=None,
                         outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page],
                         queue=False)
-
         msg.submit(conversation,
                 inputs=[qa_chain, msg, chatbot],
                 outputs=[qa_chain, msg, chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page],
@@ -231,4 +195,6 @@ def demo():
                         queue=False)
     
     demo.queue().launch(debug=True)
-    
+
+if __name__ == "__main__":
+    demo()