Update app.py

app.py

@@ -9,53 +9,64 @@ from langchain_groq import ChatGroq
 from langchain_core.prompts import PromptTemplate
 from langchain_classic.chains import RetrievalQA
 
+# Suppress unnecessary logs
 warnings.filterwarnings("ignore")
 
 # --- CONFIGURATION ---
-# On Hugging Face, this will look for a Secret named 'MY_GROQ_KEY'
+# Ensure you have 'MY_GROQ_KEY' in your HF Space Secrets
 GROQ_API_KEY = os.environ.get("MY_GROQ_KEY")
 
-# High-quality embedding model
+# High-quality embedding model (Runs on CPU)
 embeddings = HuggingFaceEmbeddings(model_name="BAAI/bge-small-en-v1.5")
 rag_chain = None
 
 def build_rag_system(file):
     global rag_chain
     if file is None: return "❌ Error: No document uploaded."
-    if not GROQ_API_KEY: return "❌ Error: Groq API Key not found in Environment Secrets."
+    if not GROQ_API_KEY: return "❌ Error: Groq API Key missing in Secrets!"
 
     try:
+        # Load PDF or TXT
         loader = PyPDFLoader(file.name) if file.name.endswith(".pdf") else TextLoader(file.name)
         documents = loader.load()
         
+        # Split into chunks
         text_splitter = RecursiveCharacterTextSplitter(chunk_size=800, chunk_overlap=80)
         texts = text_splitter.split_documents(documents)
         
+        # Create Vector Store
         vector_db = FAISS.from_documents(texts, embeddings)
         retriever = vector_db.as_retriever(search_kwargs={"k": 3})
         
+        # Initialize Groq Llama 3.3
         llm = ChatGroq(
             groq_api_key=GROQ_API_KEY, 
             model_name="llama-3.3-70b-versatile", 
             temperature=0
         )
         
-        template = """You are a professional research assistant. Answer ONLY using the context below. 
-        If the information is missing, strictly say: "I don't have enough information about this in the provided documents."
+        # Strict Prompting
+        template = """You are a professional assistant. Answer ONLY using the context. 
+        If the answer is not there, say: "I don't have enough information about this in the provided documents."
         
         Context: {context}
         Question: {question}
-        Professional Answer:"""
+        Answer:"""
         
         QA_PROMPT = PromptTemplate.from_template(template)
-        rag_chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever, chain_type_kwargs={"prompt": QA_PROMPT})
+        rag_chain = RetrievalQA.from_chain_type(
+            llm=llm, 
+            retriever=retriever, 
+            chain_type_kwargs={"prompt": QA_PROMPT}
+        )
         
-        return "✅ Document Indexed Successfully!"
+        return "✅ Document Vault Successfully Built!"
     except Exception as e:
         return f"❌ System Error: {str(e)}"
 
 def predict(message, history):
-    if rag_chain is None: return "Please upload a file and build the knowledge base first."
+    if rag_chain is None: 
+        return "Please upload a document and click Build first."
     try:
         res = rag_chain.invoke({"query": message})
         return res["result"]
@@ -63,46 +74,23 @@ def predict(message, history):
         return f"🚨 API ERROR: {str(e)}"
 
 # --- PROFESSIONAL UI DESIGN ---
-theme = gr.themes.Monochrome(
-    primary_hue="blue",
-    secondary_hue="slate",
-    radius_size="lg",
-    font=[gr.themes.GoogleFont("Inter"), "ui-sans-serif", "system-ui", "sans-serif"],
-).set(
-    button_primary_background_fill="*primary_600",
-    button_primary_background_fill_hover="*primary_700",
-    button_primary_text_color="white",
-    body_background_fill="*neutral_50",
-)
-
-with gr.Blocks(theme=theme, title="VerityVault AI") as demo:
-    with gr.Column(elem_id="container"):
-        gr.Markdown(
-            f"""
-            # 🛡️ VerityVault AI
-            ### Developed by: **Your Name**
-            *A high-precision Retrieval-Augmented Generation system for secure document analysis.*
-            """
-        )
-        
-        with gr.Row():
-            with gr.Column(scale=1):
-                file_input = gr.File(label="📄 Source Document", file_types=[".pdf", ".txt"])
-                build_btn = gr.Button("🚀 BUILD KNOWLEDGE BASE", variant="primary")
-                status = gr.Textbox(label="System Intelligence Status", placeholder="Ready for upload...", interactive=False)
-                
-                with gr.Accordion("ℹ️ Instructions", open=False):
-                    gr.Markdown("1. Upload a PDF or TXT file.\n2. Click Build.\n3. Ask questions based ONLY on that file.")
-
-            with gr.Column(scale=2):
-                gr.ChatInterface(
-                    fn=predict,
-                    type="messages",
-                    description="The engine will refuse to answer if data is not in the source file."
-                )
-        
-        gr.Markdown("---")
-        gr.Markdown("⚖️ *Powered by Llama 3.3 & Groq LPUs. Data is processed locally in volatile memory.*")
+# Removed 'type="messages"' to fix the deployment error
+with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue"), title="VerityVault AI") as demo:
+    gr.Markdown("# 🛡️ VerityVault AI")
+    gr.Markdown("### Developed by: **Bilal**")
+    gr.Markdown("*Secure document intelligence with zero hallucinations.*")
+    
+    with gr.Row():
+        with gr.Column(scale=1):
+            file_input = gr.File(label="📄 Source Document", file_types=[".pdf", ".txt"])
+            build_btn = gr.Button("🚀 BUILD VAULT", variant="primary")
+            status = gr.Textbox(label="Vault Status", interactive=False)
+            
+        with gr.Column(scale=2):
+            gr.ChatInterface(
+                fn=predict,
+                description="The vault will only answer based on your uploaded file."
+            )
 
 build_btn.click(build_rag_system, inputs=[file_input], outputs=[status])