Update app.py

app.py

@@ -3,7 +3,6 @@ import os
 import re
 import joblib
 import numpy as np
-import pandas as pd
 import gradio as gr
 
 # -------------------------
@@ -52,7 +51,6 @@ tfidf, tfidf_path = try_load(tfidf_candidates)
 logreg, logreg_path = try_load(logreg_candidates)
 lgbm, lgbm_path = try_load(lgbm_candidates)
 
-# Fallback label order (common mapping)
 DEFAULT_LABELS = ['negative', 'neutral', 'positive']
 
 # -------------------------
@@ -61,8 +59,7 @@ DEFAULT_LABELS = ['negative', 'neutral', 'positive']
 def clean_text(t):
     if t is None:
         return ""
-    s = str(t)
-    s = s.lower()
+    s = str(t).lower()
     s = re.sub(r"\s+", " ", s)
     s = re.sub(r"[^a-z0-9\s']", " ", s)
     return s.strip()
@@ -74,10 +71,8 @@ import warnings
 warnings.filterwarnings("ignore")
 
 def get_model_classes(model):
-    # some models expose .classes_, some don't
     if hasattr(model, "classes_"):
         return list(model.classes_)
-    # LightGBM might store classes_ as np.array in classifier
     if hasattr(model, "classes"):
         return list(model.classes)
     return DEFAULT_LABELS
@@ -102,14 +97,12 @@ def predict_one(text, model_choice="Logistic Regression"):
             probs = logreg.predict_proba(X)[0]
             classes = get_model_classes(logreg)
         elif model_choice == "LightGBM" and lgbm is not None:
-            # LightGBM may want dense arrays in some configs; try both
             try:
                 probs = lgbm.predict_proba(X)[0]
             except Exception:
                 probs = lgbm.predict_proba(X.toarray())[0]
             classes = get_model_classes(lgbm)
         else:
-            # fallback to whichever model exists
             if logreg is not None:
                 probs = logreg.predict_proba(X)[0]; classes = get_model_classes(logreg)
             elif lgbm is not None:
@@ -121,17 +114,14 @@ def predict_one(text, model_choice="Logistic Regression"):
     except Exception as e:
         return {"label": None, "confidence": 0.0, "html": "", "error": f"Prediction error: {e}"}
 
-    # Ensure classes + probs align
-    # If classes are not sorted in expected order, we will display them as the model provides.
     idx = int(np.argmax(probs))
     label = classes[idx]
     confidence = float(probs[idx])
 
-    # Build colored HTML bars for probabilities
     colors = {
-        'positive': '#16a34a',   # green
-        'neutral':  '#f59e0b',   # amber
-        'negative': '#ef4444'    # red
+        'positive': '#16a34a',
+        'neutral':  '#f59e0b',
+        'negative': '#ef4444'
     }
     bars_html = ""
     for c, p in zip(classes, probs):
@@ -139,8 +129,8 @@ def predict_one(text, model_choice="Logistic Regression"):
         pct = float(p) * 100.0
         bars_html += f"""
         <div style="display:flex;align-items:center;margin-bottom:8px;">
-          <div style="width:95px;font-weight:600;color:#111;">{c}</div>
-          <div style="flex:1;margin-left:10px;background:#f1f5f9;border-radius:999px;padding:3px;">
+          <div style="width:95px;font-weight:600;color:#e5e7eb;">{c}</div>
+          <div style="flex:1;margin-left:10px;background:#1f2937;border-radius:999px;padding:3px;">
             <div style="width:{pct:.2f}%;background:{col};padding:6px 10px;border-radius:999px;color:white;font-weight:700;text-align:right;">
               {pct:.1f}%
             </div>
@@ -150,7 +140,7 @@ def predict_one(text, model_choice="Logistic Regression"):
 
     header_html = f"""
     <div style="display:flex;align-items:center;gap:12px;">
-      <div style="font-size:16px;font-weight:700;">Prediction:</div>
+      <div style="font-size:16px;font-weight:700;color:#f3f4f6;">Prediction:</div>
       <div style="padding:6px 12px;border-radius:999px;background:{colors.get(str(label).lower(),'#3b82f6')};color:white;font-weight:800;">
         {label.upper()} ({confidence:.2f})
       </div>
@@ -161,36 +151,33 @@ def predict_one(text, model_choice="Logistic Regression"):
     return {"label": label, "confidence": float(confidence), "html": header_html, "error": None}
 
 # -------------------------
-# Gradio UI
+# Dark theme CSS
 # -------------------------
 css = """
-/* page background */
-body { background: linear-gradient(135deg,#fdfbfb 0%,#ebf8ff 100%); }
-
-/* central card */
+body { background: linear-gradient(135deg,#0f172a 0%,#1e293b 100%); color:#e5e7eb; }
 .app-card { 
   border-radius: 12px; 
   padding: 18px; 
-  box-shadow: 0 10px 25px rgba(11, 20, 41, 0.06); 
-  background: linear-gradient(180deg, rgba(255,255,255,0.9), rgba(255,255,255,0.82));
+  box-shadow: 0 10px 25px rgba(0, 0, 0, 0.6); 
+  background: rgba(31,41,55,0.9);
+  color: #f9fafb;
 }
-
-/* title */
 .title {
   font-weight: 800;
   font-size: 22px;
   margin-bottom: 6px;
+  color: #f9fafb;
 }
 .subtitle {
-  color: #374151;
+  color: #9ca3af;
   margin-bottom: 12px;
 }
-
-/* button */
 .gr-button {
   border-radius: 10px;
   padding: 10px 16px;
   font-weight:700;
+  background: #3b82f6 !important;
+  color: white !important;
 }
 """
 
@@ -203,33 +190,32 @@ examples = [
 with gr.Blocks(css=css, theme=gr.themes.Base()) as demo:
     with gr.Row():
         with gr.Column(scale=2):
-            gr.HTML("<div class='app-card'><div class='title'>SvaraAI — Reply Classifier</div>"
-                    "<div class='subtitle'>Paste an email reply below to classify it as <b>positive</b> / <b>neutral</b> / <b>negative</b>.</div>"
+            gr.HTML("<div class='app-card'><div class='title'>🌙 SvaraAI — Reply Classifier</div>"
+                    "<div class='subtitle'>Classify replies as <b style='color:#16a34a'>positive</b> / <b style='color:#f59e0b'>neutral</b> / <b style='color:#ef4444'>negative</b>.</div>"
                     "</div>")
-            inp = gr.Textbox(lines=5, placeholder="e.g. Thanks — let's schedule a demo next Tuesday at 10am.", label="Reply text")
-            model_choice = gr.Dropdown(choices=["Logistic Regression", "LightGBM"], value="Logistic Regression", label="Model (choose one)")
+            inp = gr.Textbox(lines=5, placeholder="Type your reply here...", label="Reply text")
+            model_choice = gr.Dropdown(choices=["Logistic Regression", "LightGBM"], value="Logistic Regression", label="Model")
             with gr.Row():
-                btn = gr.Button("Classify", variant="primary")
-                clear = gr.Button("Clear")
+                btn = gr.Button("🚀 Classify", variant="primary")
+                clear = gr.Button("🧹 Clear")
             output_label = gr.Markdown(value="**Prediction:** _waiting for input_", label="Result")
-            output_html = gr.HTML("<i>Probabilities will appear here</i>")
+            output_html = gr.HTML("<i style='color:#9ca3af;'>Probabilities will appear here</i>")
             error_box = gr.Textbox(interactive=False, visible=False)
             gr.Examples(examples=examples, inputs=[inp, model_choice], label="Try these examples")
         with gr.Column(scale=1):
-            gr.HTML("<div class='app-card'><div style='font-weight:800;margin-bottom:8px'>About</div>"
-                    "<div style='font-size:13px;color:#374151'>This demo uses a TF-IDF vectorizer and a saved classifier (Logistic Regression / LightGBM). "
+            gr.HTML("<div class='app-card'><div style='font-weight:800;margin-bottom:8px'>ℹ️ About</div>"
+                    "<div style='font-size:13px;color:#d1d5db'>This demo uses a TF-IDF vectorizer and a saved classifier (Logistic Regression / LightGBM). "
                     "Upload your saved pickles to <code>models/</code> as described in README.md.</div></div>")
-            # small quick test panel
             stats_md = gr.Markdown("**Model files detected:**<br>"
                                    f"- TF-IDF: `{tfidf_path or 'NOT FOUND'}`  \n"
                                    f"- LogReg: `{logreg_path or 'NOT FOUND'}`  \n"
                                    f"- LGBM: `{lgbm_path or 'NOT FOUND'}`  \n")
-            download_note = gr.Markdown("<small>If a model is missing upload it to <code>models/</code> or rename files appropriately.</small>")
+            download_note = gr.Markdown("<small style='color:#9ca3af;'>If a model is missing upload it to <code>models/</code> or rename files appropriately.</small>")
 
     def run_and_format(text, model_choice):
         res = predict_one(text, model_choice)
         if res.get("error"):
-            return f"**Error:** {res['error']}", "", gr.update(value=f"<div style='color:#b91c1c;font-weight:700'>{res['error']}</div>")
+            return f"**Error:** {res['error']}", "", gr.update(value=f"<div style='color:#ef4444;font-weight:700'>{res['error']}</div>")
         label = res["label"]
         conf = res["confidence"]
         html = res["html"]
@@ -237,10 +223,9 @@ with gr.Blocks(css=css, theme=gr.themes.Base()) as demo:
         return md, str(round(conf, 3)), gr.update(value=html)
 
     btn.click(run_and_format, inputs=[inp, model_choice], outputs=[output_label, error_box, output_html])
-    clear.click(lambda: ("**Prediction:** _waiting for input_", "", gr.update(value="<i>Probabilities will appear here</i>")), [], [output_label, error_box, output_html])
+    clear.click(lambda: ("**Prediction:** _waiting for input_", "", gr.update(value="<i style='color:#9ca3af;'>Probabilities will appear here</i>")), [], [output_label, error_box, output_html])
 
-    # footer
-    gr.HTML("<div style='margin-top:18px;color:#6b7280;font-size:13px'>Built for the SvaraAI assignment • Upload your model pickles into <code>models/</code></div>")
+    gr.HTML("<div style='margin-top:18px;color:#9ca3af;font-size:13px'>🌌 Built for the SvaraAI assignment • Upload your model pickles into <code>models/</code></div>")
 
 if __name__ == "__main__":
     demo.launch(server_name="0.0.0.0", server_port=int(os.environ.get("PORT", 7860)))