Upload app.py with huggingface_hub

app.py

@@ -187,9 +187,15 @@ def generate_table_html(rows):
                         <span class="text-sm font-medium text-slate-700">{row['directTime']}</span>
                     </div>
                     <div class="flex justify-between items-center">
-                        <span class="text-xs text-slate-500">Recall:</span>
+                        <span class="text-xs text-slate-500">Recall@10:</span>
                         <span class="text-xs font-mono bg-slate-100 px-1.5 rounded text-slate-600">
-                            {row['recall']}
+                            {row['recall_10']}
+                        </span>
+                    </div>
+                    <div class="flex justify-between items-center">
+                        <span class="text-xs text-slate-500">Recall@5:</span>
+                        <span class="text-xs font-mono bg-slate-100 px-1.5 rounded text-slate-600">
+                            {row['recall_5']}
                         </span>
                     </div>
                 </div>
@@ -205,11 +211,18 @@ def generate_table_html(rows):
             </td>
             <td class="px-6 py-4 whitespace-nowrap align-top border-b border-slate-100">
                 <div class="flex flex-col justify-center h-full pt-1">
+                <div class="flex flex-col justify-center h-full pt-1 space-y-2">
                     <div class="flex items-center">
-                        <span class="text-lg font-bold text-green-600">{row['efficiency']}</span>
-                        <span class="material-symbols-outlined text-green-600 text-sm ml-1">bolt</span>
+                        <span class="text-sm font-bold text-green-600">{row['efficiency_sharded']}</span>
+                        <span class="material-symbols-outlined text-green-600 text-[10px] ml-1">bolt</span>
+                    </div>
+                    <span class="text-[10px] text-green-700/60 font-semibold tracking-wide">[vs Sharded]</span>
+                    
+                    <div class="flex items-center mt-2 border-t border-slate-100 pt-2">
+                        <span class="text-sm font-bold text-slate-600">{row['efficiency_base']}</span>
+                         <span class="material-symbols-outlined text-slate-500 text-[10px] ml-1">bolt</span>
                     </div>
-                    <span class="text-[10px] text-green-700/60 uppercase font-semibold tracking-wide">Faster</span>
+                    <span class="text-[10px] text-slate-500/60 font-semibold tracking-wide">[vs Base]</span>
                 </div>
             </td>
         </tr>
@@ -339,17 +352,34 @@ def run_benchmark(query):
             prod_contents = set(p.payload.get('text', str(p.payload)) for p in prod_results)
             
             if base_contents:
-                intersection = len(base_contents.intersection(prod_contents))
-                recall = (intersection / len(base_contents)) * 100
+                intersection_10 = len(base_contents.intersection(prod_contents))
+                recall_10 = (intersection_10 / len(base_contents)) * 100
+                
+                # Recall@5
+                base_contents_5 = set(p.payload.get('text', str(p.payload)) for p in base_results[:5])
+                prod_contents_5 = set(p.payload.get('text', str(p.payload)) for p in prod_results[:5])
+                if base_contents_5:
+                    intersection_5 = len(base_contents_5.intersection(prod_contents_5))
+                    recall_5 = (intersection_5 / len(base_contents_5)) * 100
+                else:
+                    recall_5 = 0.0
             else:
-                recall = 0.0
+                recall_10 = 0.0
+                recall_5 = 0.0
                 
-            # Efficiency Gain: (Direct_Sharded - Optimized) / Direct_Sharded
-            # Using real sharded time
+            # Efficiency Gain: (Reference - Optimized) / Reference
+            
+            # 1. vs Direct Sharded
             if direct_sharded_time_ms > 0:
-                eff_gain = ((direct_sharded_time_ms - latency_ms) / direct_sharded_time_ms) * 100
+                eff_gain_sharded = ((direct_sharded_time_ms - latency_ms) / direct_sharded_time_ms) * 100
             else:
-                eff_gain = 0.0
+                eff_gain_sharded = 0.0
+
+            # 2. vs Base (No Sharding)
+            if baseline_time_ms > 0:
+                eff_gain_base = ((baseline_time_ms - latency_ms) / baseline_time_ms) * 100
+            else:
+                eff_gain_base = 0.0
             
             # Formatting
             row = {
@@ -364,8 +394,11 @@ def run_benchmark(query):
                 "confDisplay": f"{confidence*100:.1f}%",
                 "directTime": f"{direct_sharded_time_ms:.1f} ms",
                 "baselineTime": f"{baseline_time_ms:.1f} ms",
-                "recall": f"{recall:.1f}%",
-                "efficiency": f"{eff_gain:.1f}%"
+                "baselineTime": f"{baseline_time_ms:.1f} ms",
+                "recall_10": f"{recall_10:.1f}%",
+                "recall_5": f"{recall_5:.1f}%",
+                "efficiency_sharded": f"{eff_gain_sharded:.1f}%",
+                "efficiency_base": f"{eff_gain_base:.1f}%"
             }
             rows.append(row)