Update app/app.py

app/app.py

@@ -1,120 +1,124 @@
-# ----------------------------
-# FILE: app.py (FINAL CLEAN VERSION)
-# ----------------------------
-import os
-import streamlit as st
-
-# Avoid TensorFlow warnings
-os.environ.setdefault("TRANSFORMERS_NO_TF", "1")
-os.environ["STREAMLIT_SERVER_HEADLESS"] = "true"   # Required for HuggingFace Spaces
-
-# ----------------------------------------------------
-# MAIN PAGE CONFIG (ONLY HERE) — required by Streamlit
-# ----------------------------------------------------
-st.set_page_config(
-    page_title="AI Customer Feedback Analyzer — SaaS Pro",
-    page_icon="🧠",
-    layout="wide"
-)
-
-# ----------------------------------------------------
-# GLOBAL SESSION STATE
-# ----------------------------------------------------
-if "logged_in" not in st.session_state:
-    st.session_state.logged_in = False
-
-if "username" not in st.session_state:
-    st.session_state.username = ""
-
-# ----------------------------------------------------
-# PREMIUM SAAS LANDING HEADER  ✔ FIXED HTML, NO RAW TEXT
-# ----------------------------------------------------
-st.markdown(
-    """
-<div style='background:linear-gradient(90deg,#00c2b8,#007bff);
-            padding:22px;border-radius:12px;margin-bottom:15px; color:white;'>
-
-  <h1 style='margin:0;'>🧠 AI Customer Feedback Analyzer</h1>
-  <p style='opacity:0.9;'>Hybrid BERT + RoBERTa • Explainable AI (LIME) • Premium SaaS UI</p>
-
-  <div style='text-align:right; margin-top:-40px;'>
-      <strong>Built by Lalit Chaudhari</strong><br>
-      <a href='https://github.com/LalitChaudhari851' style='color:white;text-decoration:none;'>GitHub</a> •
-      <a href='https://www.linkedin.com/in/lalit-chaudhari-9a62b8190/'
-         style='color:white;text-decoration:none;'>LinkedIn</a>
-  </div>
-
-</div>
-""",
-    unsafe_allow_html=True
-)
-
-# ----------------------------------------------------
-# CTA — open login page
-# ----------------------------------------------------
-st.markdown("### 👇 Start using the app")
-
-if st.button("🔐 Open Login Page"):
-    st.switch_page("pages/login.py")
-
-st.markdown("---")
-
-# ----------------------------------------------------
-# SHARE MODELS WITH PAGES (CACHE = FAST)
-# ----------------------------------------------------
-@st.cache_resource(show_spinner=True)
-def get_shared_resources():
-    """Load ML models only once, share across pages."""
-    resources = {
-        "bert": None,
-        "roberta": None,
-        "has_transformers": False,
-        "has_wordcloud": False,
-        "has_lime": False
-    }
-
-    try:
-        from transformers import pipeline
-
-        resources["bert"] = pipeline(
-            "sentiment-analysis",
-            model="distilbert-base-uncased-finetuned-sst-2-english"
-        )
-
-        resources["roberta"] = pipeline(
-            "sentiment-analysis",
-            model="cardiffnlp/twitter-roberta-base-sentiment"
-        )
-
-        resources["has_transformers"] = True
-
-    except Exception as e:
-        print("Transformer load issue:", e)
-
-    try:
-        import wordcloud
-        resources["has_wordcloud"] = True
-    except:
-        pass
-
-    try:
-        import lime
-        resources["has_lime"] = True
-    except:
-        pass
-
-    return resources
-
-# ----------------------------------------------------
-# PRELOAD MODELS BUTTON
-# ----------------------------------------------------
-if st.button("⚡ Preload ML Models (Optional)"):
-
-    _ = get_shared_resources()
-    st.success("Models preloaded (if available)!")
-
-# ----------------------------------------------------
-# FOOTER
-# ----------------------------------------------------
-st.markdown("---")
-st.caption("🚀 AI Customer Feedback Analyzer • Hybrid BERT + RoBERTa • Premium SaaS UI")
+# ----------------------------
+# FILE: app.py
+# ----------------------------
+import os
+import streamlit as st
+
+os.environ["TRANSFORMERS_NO_TF"] = "1"
+os.environ["STREAMLIT_SERVER_HEADLESS"] = "true"
+
+# ----------------------------
+# PAGE CONFIG
+# ----------------------------
+st.set_page_config(
+    page_title="AI Customer Feedback Analyzer",
+    page_icon="🧠",
+    layout="wide"
+)
+
+# ----------------------------
+# HEADER
+# ----------------------------
+st.markdown(
+    """
+    <div style='background:linear-gradient(90deg,#00c2b8,#007bff);
+                padding:22px;border-radius:12px;color:white;'>
+        <h1>🧠 AI Customer Feedback Analyzer</h1>
+        <p>Hybrid BERT + RoBERTa • Sentiment Analysis</p>
+    </div>
+    """,
+    unsafe_allow_html=True
+)
+
+st.markdown("---")
+
+# ----------------------------
+# LOAD MODELS (CACHED)
+# ----------------------------
+@st.cache_resource(show_spinner=True)
+def load_models():
+    from transformers import pipeline
+
+    bert = pipeline(
+        "sentiment-analysis",
+        model="distilbert-base-uncased-finetuned-sst-2-english"
+    )
+
+    roberta = pipeline(
+        "sentiment-analysis",
+        model="cardiffnlp/twitter-roberta-base-sentiment"
+    )
+
+    return bert, roberta
+
+
+bert, roberta = load_models()
+
+# ----------------------------
+# ROBERTA LABEL FIX ✅
+# ----------------------------
+ROBERTA_LABEL_MAP = {
+    "LABEL_0": "Negative",
+    "LABEL_1": "Neutral",
+    "LABEL_2": "Positive"
+}
+
+# ----------------------------
+# ANALYSIS FUNCTIONS
+# ----------------------------
+def analyze_bert(text):
+    r = bert(text)[0]
+    return r["label"], round(r["score"] * 100, 2)
+
+
+def analyze_roberta(text):
+    r = roberta(text)[0]
+    sentiment = ROBERTA_LABEL_MAP.get(r["label"], "Neutral")
+    return sentiment, round(r["score"] * 100, 2)
+
+# ----------------------------
+# USER INPUT
+# ----------------------------
+st.subheader("✍️ Enter Customer Feedback")
+
+text = st.text_area(
+    "Feedback",
+    placeholder="Example: I love the product but delivery was slow"
+)
+
+# ----------------------------
+# ANALYZE
+# ----------------------------
+if st.button("🔍 Analyze Sentiment"):
+
+    if not text.strip():
+        st.warning("Please enter feedback text.")
+    else:
+        with st.spinner("Analyzing..."):
+            bert_sent, bert_conf = analyze_bert(text)
+            rob_sent, rob_conf = analyze_roberta(text)
+
+        col1, col2 = st.columns(2)
+
+        with col1:
+            st.subheader("🤖 BERT")
+            st.write(f"Sentiment: **{bert_sent}**")
+            st.write(f"Confidence: **{bert_conf}%**")
+
+        with col2:
+            st.subheader("🧠 RoBERTa")
+            st.write(f"Sentiment: **{rob_sent}**")
+            st.write(f"Confidence: **{rob_conf}%**")
+
+        st.markdown("---")
+
+        if bert_sent.lower() == rob_sent.lower():
+            st.success(f"✅ Final Verdict: **{bert_sent}**")
+        else:
+            st.info("⚠️ Final Verdict: **Mixed Sentiment**")
+
+# ----------------------------
+# FOOTER
+# ----------------------------
+st.markdown("---")
+st.caption("🚀 Built by Lalit Chaudhari")