Update app.py

app.py

@@ -13,20 +13,18 @@ from langchain_core.output_parsers import StrOutputParser
 print("Token:", os.environ.get("HUGGINGFACEHUB_API_TOKEN"))
 token = os.environ.get("HUGGINGFACEHUB_API_TOKEN")
 
+
 # ------------------------
 # LLM Model (LLaMA 2)
 # ------------------------
-# Initialize the HuggingFace text-generation pipeline
 pipe = pipeline(
     task="text-generation",
     model="meta-llama/Llama-2-7b-hf",
-    token=token,  # <- aqui você passa o token
     temperature=0.7,
     max_new_tokens=512,
     device_map="auto"
 )
 
-# Wrap the pipeline into a LangChain LLM object
 llm = HuggingFacePipeline(
     pipeline=pipe,
     model_kwargs={"temperature": 0.7}
@@ -35,7 +33,6 @@ llm = HuggingFacePipeline(
 # ------------------------
 # Prompt template
 # ------------------------
-# Define a template for asking questions based on documents
 prompt = PromptTemplate.from_template(
     """Given the following extracted parts of a long document and a question, create a final answer with references.
 If you don't know the answer, just say that you don't know.
@@ -43,72 +40,61 @@ Question: {question}"""
 )
 
 # Global variable to store the QA chain
-qa_chain = None
+simple_chain = None
 
 # ------------------------
-# Function to process URLs
+# Function to process URLs with real-time logging
 # ------------------------
-def process_urls(url1, url2, url3):
-    global qa_chain
+def process_urls_with_logs(url1, url2, url3):
+    global simple_chain
     
-    # Collect non-empty URLs
     urls = [url1, url2, url3]
     urls = [u for u in urls if u.strip() != ""]
     
     if len(urls) == 0:
-        return "⚠️ Please provide at least one URL."
+        yield "⚠️ Please provide at least one URL."
+        return
     
-    # Load documents from the URLs
+    yield "⏳ Loading URLs..."
     loader = UnstructuredURLLoader(urls=urls)
     documents = loader.load()
     
-    # Split documents into smaller chunks for embedding
+    yield "⏳ Creating the chunks..."
     text_splitter = RecursiveCharacterTextSplitter(
         chunk_size=600,
         chunk_overlap=200
     )
     splits = text_splitter.split_documents(documents)
     
-    # Create embeddings for each chunk
+    yield "⏳ Creating embeddings..."
     embeddings = HuggingFaceEmbeddings(
         model_name="mixedbread-ai/mxbai-embed-large-v1"
     )
     
-    # Build a vectorstore (FAISS) from the document embeddings
+    yield "⏳ Creating a vector database-like structure (FAISS)..."
     vectorstore = FAISS.from_documents(
         documents=splits,
         embedding=embeddings
     )
     
-    # Create a retriever to fetch relevant document chunks
+    yield "⏳ Starting LLM model (meta-llama/Llama-2-7b-hf)..."
     retriever = vectorstore.as_retriever()
     
-    # Create a simple QA chain using the prompt and LLM
     from langchain_core.runnables import RunnableSequence
     simple_chain = RunnableSequence(prompt, llm, StrOutputParser())
     
-    # Store the retriever and chain in a global variable
-    qa_chain = {"retriever": retriever, "chain": simple_chain}
-    
-    return "✅ URLs processed successfully!"
+    yield "✅ URLs processed successfully!"
 
 # ------------------------
 # Function to answer questions
 # ------------------------
 def ask_question(question):
-    global qa_chain
+    global simple_chain
 
-    if qa_chain is None:
+    if simple_chain is None:
         return "⚠️ Please process URLs first."
-
-    # Retrieve the most relevant documents for the question
-    docs = qa_chain["retriever"].get_relevant_documents(question)
     
-    # Combine document texts into a single string for the prompt
-    combined_text = "\n\n".join([d.page_content for d in docs])
-
-    # Run the QA chain with the question and context
-    result = qa_chain["chain"].invoke({"question": question, "context": combined_text})
+    result = simple_chain.invoke({"question": question})
     return result
 
 # ------------------------
@@ -116,7 +102,6 @@ def ask_question(question):
 # ------------------------
 with gr.Blocks() as app:
     with gr.Row():
-        
         # Sidebar: URL input and processing
         with gr.Column(scale=1):
             gr.Markdown("## 📌 Insert URLs")
@@ -126,8 +111,8 @@ with gr.Blocks() as app:
             url3 = gr.Textbox(label="URL 3")
             
             process_btn = gr.Button("Process URLs")
-            status_output = gr.Textbox(label="Status")
-
+            status_output = gr.Textbox(label="Status", lines=8)
+        
         # Main Area: Question input and answer output
         with gr.Column(scale=2):
             gr.Markdown("## ✍️ Write your question")
@@ -140,14 +125,15 @@ with gr.Blocks() as app:
             
             ask_btn = gr.Button("Ask")
             answer_output = gr.Textbox(label="Answer", lines=8)
-
+    
     # Connect buttons to their functions
     process_btn.click(
-        process_urls,
+        process_urls_with_logs,
         inputs=[url1, url2, url3],
-        outputs=status_output
+        outputs=status_output,
+        streaming=True  # ⚡️ atualiza logs em tempo real
     )
-
+    
     ask_btn.click(
         ask_question,
         inputs=question_box,