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
@@ -15,14 +14,16 @@ from langchain.memory import ConversationBufferMemory
 from langchain_community.llms import HuggingFaceEndpoint
 import torch
 
-list_llm = ["meta-llama/Meta-Llama-3-8B-Instruct", "mistralai/Mistral-7B-Instruct-v0.2"]  
+# Added Deepseek model to the list
+list_llm = [
+    "meta-llama/Meta-Llama-3-8B-Instruct", 
+    "mistralai/Mistral-7B-Instruct-v0.2",
+    "deepseek-ai/deepseek-llm-7b-chat"
+]  
 list_llm_simple = [os.path.basename(llm) for llm in list_llm]
 
-# Load and split PDF document
+# Rest of the functions remain the same until demo()
 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:
@@ -34,36 +35,20 @@ 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
-from langchain_community.llms import HuggingFaceEndpoint
-
-# 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(
-            repo_id=llm_model,
-            huggingfacehub_api_token=api_token,
-            temperature=temperature,
-            max_new_tokens=max_tokens,
-            top_k=top_k,
-            task="text-generation"  # Explicitly specify the task type
-        )
-    else:
-        llm = HuggingFaceEndpoint(
-            huggingfacehub_api_token=api_token,
-            repo_id=llm_model,
-            temperature=temperature,
-            max_new_tokens=max_tokens,
-            top_k=top_k,
-            task="text-generation"  # Explicitly specify the task type
-        )
+    llm = HuggingFaceEndpoint(
+        repo_id=llm_model,
+        huggingfacehub_api_token=api_token,
+        temperature=temperature,
+        max_new_tokens=max_tokens,
+        top_k=top_k,
+        task="text-generation"
+    )
     
     memory = ConversationBufferMemory(
         memory_key="chat_history",
@@ -81,37 +66,28 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, pr
         verbose=False,
     )
     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:
@@ -120,14 +96,11 @@ 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 = []
@@ -136,10 +109,15 @@ def upload_file(file_obj):
         list_file_path.append(file_path)
     return list_file_path
 
-
 def demo():
-    # with gr.Blocks(theme=gr.themes.Default(primary_hue="sky")) as demo:
-    with gr.Blocks(theme=gr.themes.Default(primary_hue="red", secondary_hue="pink", neutral_hue = "sky")) as demo:
+    # Modified theme to use dark blue colors
+    theme = gr.themes.Default(
+        primary_hue="indigo",
+        secondary_hue="blue",
+        neutral_hue="slate"
+    )
+    
+    with gr.Blocks(theme=theme) as demo:
         vector_db = gr.State()
         qa_chain = gr.State()
         gr.HTML("<center><h1>RAG PDF chatbot</h1><center>")
@@ -154,10 +132,10 @@ def demo():
                 with gr.Row():
                     db_btn = gr.Button("Create vector database")
                 with gr.Row():
-                        db_progress = gr.Textbox(value="Not initialized", show_label=False) # label="Vector database status", 
+                    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") # info="Select LLM", show_label=False
+                    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():
@@ -165,11 +143,11 @@ def demo():
                         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)
+                            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) # label="Chatbot status", 
+                    llm_progress = gr.Textbox(value="Not initialized", show_label=False)
 
             with gr.Column(scale = 200):
                 gr.Markdown("<b>Step 2 - Chat with your Document</b>")
@@ -190,32 +168,31 @@ def demo():
                     submit_btn = gr.Button("Submit")
                     clear_btn = gr.ClearButton([msg, chatbot], value="Clear")
             
-        # Preprocessing events
-        db_btn.click(initialize_database, \
-            inputs=[document], \
+        # Event handlers
+        db_btn.click(initialize_database, 
+            inputs=[document], 
             outputs=[vector_db, db_progress])
-        qachain_btn.click(initialize_LLM, \
-            inputs=[llm_btn, slider_temperature, slider_maxtokens, slider_topk, vector_db], \
-            outputs=[qa_chain, llm_progress]).then(lambda:[None,"",0,"",0,"",0], \
-            inputs=None, \
-            outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page], \
+        qachain_btn.click(initialize_LLM, 
+            inputs=[llm_btn, slider_temperature, slider_maxtokens, slider_topk, vector_db], 
+            outputs=[qa_chain, llm_progress]).then(lambda:[None,"",0,"",0,"",0], 
+            inputs=None, 
+            outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page], 
             queue=False)
 
-        # Chatbot events
-        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], \
+        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], 
             queue=False)
-        submit_btn.click(conversation, \
-            inputs=[qa_chain, msg, chatbot], \
-            outputs=[qa_chain, msg, chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page], \
+        submit_btn.click(conversation, 
+            inputs=[qa_chain, msg, chatbot], 
+            outputs=[qa_chain, msg, chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page], 
             queue=False)
-        clear_btn.click(lambda:[None,"",0,"",0,"",0], \
-            inputs=None, \
-            outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page], \
+        clear_btn.click(lambda:[None,"",0,"",0,"",0], 
+            inputs=None, 
+            outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page], 
             queue=False)
-    demo.queue().launch(debug=True)
 
+    demo.queue().launch(debug=True)
 
 if __name__ == "__main__":
     demo()