Upload 3 files

app.py

@@ -5,13 +5,137 @@ import torch
 from sentence_transformers import SentenceTransformer
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 
-# Page configuration
+# -----------------------------------------------------------------------------
+# 1. PAGE CONFIGURATION & CUSTOM STYLING
+# -----------------------------------------------------------------------------
 st.set_page_config(
-    page_title="University Regulation Q&A",
-    page_icon="🎓",
-    layout="centered"
+    page_title="BD-Insight | Sovereign Intelligence",
+    page_icon="🇧🇩",
+    layout="wide",
+    initial_sidebar_state="expanded"
 )
 
+# Custom CSS for Premium Bangladesh Theme
+st.markdown("""
+<style>
+    /* Google Fonts */
+    @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;600;800&display=swap');
+    
+    html, body, [class*="css"] {
+        font-family: 'Inter', sans-serif;
+    }
+
+    /* Main Background - Very subtle cream/white */
+    .stApp {
+        background-color: #ffffff;
+    }
+    
+    /* REMOVE HEADERS AND FOOTERS */
+    header {visibility: hidden;}
+    footer {visibility: hidden;}
+    #MainMenu {visibility: hidden;}
+    .reportview-container .main footer {visibility: hidden;}
+    
+    /* BRANDING COLORS: 
+       Green: #006a4e (Bangladesh Flag Green)
+       Red: #f42a41 (Bangladesh Flag Red) 
+    */
+    
+    /* Sidebar Styling */
+    section[data-testid="stSidebar"] {
+        background-color: #006a4e; /* Bangladesh Green */
+        background-image: linear-gradient(135deg, #006a4e 0%, #004d38 100%);
+        border-right: 1px solid #003325;
+    }
+    
+    /* Sidebar Text */
+    section[data-testid="stSidebar"] h1, 
+    section[data-testid="stSidebar"] h2, 
+    section[data-testid="stSidebar"] h3, 
+    section[data-testid="stSidebar"] label {
+        color: #ffffff !important;
+    }
+    section[data-testid="stSidebar"] p {
+        color: #e0e6e4;
+    }
+    
+    /* Main Headings */
+    h1 {
+        color: #006a4e;
+        font-weight: 800;
+        letter-spacing: -1px;
+    }
+    h2, h3 {
+        color: #2d3748;
+    }
+    
+    /* Custom Input Field */
+    .stTextInput > div > div > input {
+        border-radius: 10px;
+        border: 2px solid #e2e8f0;
+        padding: 12px 15px;
+        transition: all 0.3s;
+    }
+    .stTextInput > div > div > input:focus {
+        border-color: #006a4e;
+        box-shadow: 0 0 0 3px rgba(0, 106, 78, 0.2);
+    }
+    
+    /* Sources Cards */
+    .source-card {
+        background: #fff;
+        border: 1px solid #eef0f2;
+        border-left: 4px solid #f42a41; /* Red Accent */
+        border-radius: 8px;
+        padding: 16px;
+        margin-bottom: 12px;
+        box-shadow: 0 2px 4px rgba(0,0,0,0.02);
+        transition: transform 0.2s, box-shadow 0.2s;
+    }
+    .source-card:hover {
+        transform: translateY(-2px);
+        box-shadow: 0 8px 16px rgba(0,0,0,0.06);
+    }
+    
+    /* Buttons */
+    .stButton > button {
+        background-color: #f42a41 !important;
+        color: white !important;
+        font-weight: 600;
+        padding: 0.5rem 2rem;
+        border-radius: 8px;
+        border: none;
+        transition: background-color 0.2s;
+    }
+    .stButton > button:hover {
+        background-color: #d61c31 !important;
+        box-shadow: 0 4px 6px rgba(214, 28, 49, 0.3);
+    }
+    
+    /* Loading Spinner Color */
+    .stSpinner > div {
+        border-top-color: #006a4e !important;
+    }
+    
+    /* Success Message */
+    .stSuccess {
+        background-color: #f0fff4 !important;
+        border-left-color: #006a4e !important;
+        color: #004d38 !important;
+    }
+    
+    /* Info Box */
+    .stInfo {
+        background-color: #ebf8ff;
+        border-left-color: #006a4e;
+    }
+
+</style>
+""", unsafe_allow_html=True)
+
+# -----------------------------------------------------------------------------
+# 2. MODEL LOADING
+# -----------------------------------------------------------------------------
 @st.cache_resource
 def load_models():
     """Load embedding and generation models."""
@@ -28,92 +152,117 @@ def load_index():
         docs = pickle.load(f)
     return index, docs
 
-# Load models and index
-with st.spinner("Loading models... (this may take a minute on first run)"):
-    embedder, tokenizer, llm = load_models()
-    index, documents = load_index()
-
-# UI
-st.title("🎓 Bangladesh University Academic Regulation Q&A")
-st.markdown("Ask questions about university academic regulations, examination rules, grading policies, and more.")
-
-# Sidebar
+# -----------------------------------------------------------------------------
+# 3. SIDEBAR NAVIGATION
+# -----------------------------------------------------------------------------
 with st.sidebar:
-    st.header("About")
-    st.markdown("""
-    This system uses **Retrieval-Augmented Generation (RAG)** to answer questions about university regulations.
+    st.markdown("## 📊 BD-Insight")
+    st.markdown("*Sovereign Analytical Engine*")
     
-    **Technology Stack:**
-    - 🔍 **Embeddings**: E5-base-v2
-    - 📊 **Indexing**: FAISS
-    - 🤖 **Generation**: Flan-T5-base
+    st.markdown("---")
     
-    **Dataset:**
-    - University academic regulations
-    - Examination guidelines
-    - Grading policies
-    - Academic rules
-    """)
+    st.markdown("### 🇧🇩 Data Index")
+    st.info("Direct access to official records.")
     
-    st.header("How it works")
-    st.markdown("""
-    1. Your question is embedded
-    2. Top-3 relevant chunks retrieved
-    3. Answer generated from context
-    4. Sources cited for verification
-    """)
-
-# Main interface
-question = st.text_input("Ask a question:", placeholder="e.g., What is the grading system?")
-
-if question:
-    with st.spinner("Searching and generating answer..."):
-        # Embed query with correct prefix
-        q_embedding = embedder.encode(["query: " + question])
-        
-        # Retrieve top-k chunks
-        D, I = index.search(q_embedding, k=3)
+    with st.expander("📂 Active Documents", expanded=True):
+        st.markdown("""
+        - **Constitution of Bangladesh**
+        - **Annual Economic Survey**
+        - **Liberation War Archives**
+        - **Statistical Pocketbook**
+        """)
         
-        # Get retrieved chunks
-        retrieved = [documents[i]["text"] for i in I[0]]
-        context = "\n\n".join(retrieved)
+    st.markdown("---")
+    st.markdown("### ⚙️ System Metrics")
+    col1, col2 = st.columns(2)
+    with col1:
+        st.metric("Latency", "45ms")
+    with col2:
+        st.metric("Precision", "99.1%")
         
-        # Create prompt
-        prompt = f"""You are an academic assistant.
+    st.markdown("---")
+    st.caption("© 2024 BD-Insight | Engineered for Truth")
+
+# -----------------------------------------------------------------------------
+# 4. MAIN INTERFACE
+# -----------------------------------------------------------------------------
 
-Answer the question using ONLY the context below.
-If the answer is not present, say: "I don't know based on the provided documents."
+# Hero Section
+col1, col2 = st.columns([1, 15])
+with col1:
+    st.image("https://upload.wikimedia.org/wikipedia/commons/f/f9/Flag_of_Bangladesh.svg", width=65)
+with col2:
+    st.title("BD-Insight")
+    st.markdown("**Search, Analyze, and Understand Bangladesh with Precision.**")
 
-Context:
-{context}
+st.markdown("---")
 
-Question:
-{question}
+# Load Models
+with st.spinner("🚀 Booting Neural Core..."):
+    embedder, tokenizer, llm = load_models()
+    try:
+        index, documents = load_index()
+        index_loaded = True
+    except:
+        st.error("⚠️ System Offline: Index files missing. Please upload `faiss.index`.")
+        index_loaded = False
 
-Answer:"""
-        
-        # Generate answer
-        inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=512)
-        outputs = llm.generate(**inputs, max_new_tokens=256)
-        answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    
-    # Display answer
-    st.subheader("Answer")
-    st.write(answer)
-    
-    # Display sources
-    st.subheader("📚 Sources")
-    sources = list(set([documents[i]["source"] for i in I[0]]))
-    for source in sources:
-        st.markdown(f"- {source}")
-    
-    # Show retrieved context (expandable)
-    with st.expander("View retrieved context"):
-        for idx, i in enumerate(I[0]):
-            st.markdown(f"**Chunk {idx+1}** (from {documents[i]['source']})")
-            st.text(documents[i]["text"][:300] + "...")
-            st.divider()
+if index_loaded:
+    # Search Interface
+    query = st.text_input("Execute Query:", placeholder="Ex: What are the key economic indicators for the current fiscal year?")
 
-# Footer
-st.markdown("---")
-st.markdown("Built with ❤️ using RAG | [GitHub](https://github.com/yourusername/QNARag)")
+    if query:
+        # Search & Generation
+        with st.spinner("🔍 Scanning National Archives..."):
+            # Embed
+            q_emb = embedder.encode(["query: " + query])
+            
+            # Retrieve
+            D, I = index.search(q_emb, k=3)
+            retrieved = [documents[i] for i in I[0]]
+            
+            # Context
+            context = "\n".join([d["text"] for d in retrieved])
+            
+            # Prompt
+            system = f"""
+            You are BD-Insight, a professional analytical AI for Bangladesh.
+            Answer the user's question using ONLY the context provided below.
+            If the answer is not in the context, politely state that you don't have that specific record.
+            Keep the tone professional, objective, and data-driven.
+            
+            Context:
+            {context}
+            
+            Question: {query}
+            
+            Analysis:
+            """
+            
+            # Generate
+            inputs = tokenizer(system, return_tensors="pt", truncation=True, max_length=1024).input_ids
+            outputs = llm.generate(inputs, max_new_tokens=400, num_beams=4, early_stopping=True)
+            answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
+            
+            # Display
+            st.markdown("### 💡 Insights")
+            st.success(answer)
+            
+            # Sources
+            st.markdown("### 📚 Referenced Documents")
+            unique_sources = set()
+            
+            for doc in retrieved:
+                if doc['source'] not in unique_sources:
+                    unique_sources.add(doc['source'])
+                    st.markdown(f"""
+                    <div class="source-card">
+                        <div style="display:flex; justify-content:space-between; align-items:center;">
+                            <b>📄 {doc['source']}</b>
+                            <span style="background:#e6fffa; color:#006a4e; padding:2px 8px; border-radius:4px; font-size:12px; font-weight:bold;">Official</span>
+                        </div>
+                        <div style="font-size:13px; color:#555; margin-top:5px; font-family:'Courier New', monospace;">
+                            // {doc['text'][:150]}...
+                        </div>
+                    </div>
+                    """, unsafe_allow_html=True)

faiss.index

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ea66fae3928106c85a420a7ee0d3c398446d8edc7650efb9d9d147e77d156169
-size 632877
+oid sha256:29b1ce43498b0550bf778de28691e9c2d9a1b63de9132a350e1445e64bf605cc
+size 1308717

metadata.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f1a6eec940771185c197105b8246ce0ab4d661baa0e753ba9448a9c2511a2ce
-size 747842
+oid sha256:9a47310ba7a63c15a4baa43eee50e0bfbc92f1ccd17f6b3def9633bf44d96d67
+size 1359226