NN view: one per row with prompt, top to bottom

app.py

@@ -14,16 +14,18 @@ TOP1_ID = SAMPLE_IDS[0] if SAMPLE_IDS else None
 
 
 def build_nn_view(sample_id: str | None):
-    """NN view: NN1-NN10 from baseline (in prompt order). Each has spec + m_wav."""
+    """NN view: NN1-NN10 from baseline (in prompt order). Each row: prompt + spec + m_wav."""
     if not sample_id:
-        return (None,) * (10 * 2)
+        return (None,) * (10 + 10 * 2)
     data = get_nn_demo_paths(sample_id, top_k=10)
     out = []
-    for nn in data.get("nn_list", [])[:10]:
+    for i, nn in enumerate(data.get("nn_list", [])[:10]):
+        prompt = nn.get("prompt", "") or ""
+        out.append(f"**NN{i+1}:** {prompt}" if prompt else "")
         out.extend([nn.get("spec"), nn.get("m_wav")])
-    while len(out) < 20:
+    while len(out) < 30:
         out.append(None)
-    return tuple(out[:20])
+    return tuple(out[:30])
 
 
 def build_results_view(sample_id: str | None):
@@ -61,16 +63,17 @@ with gr.Blocks(title="NearestNeighbor Audio Demo", css=".gradio-container { max-
 **Audio labels**: **BG** = background noise | **FG** = generated foreground | **Mix** = BG + FG
 """)
 
-    # ---- NN View: NN1-NN10 from baseline (in prompt order) ----
+    # ---- NN View: NN1-NN10 from baseline, one per row with prompt ----
     with gr.Column(visible=True) as nn_col:
         gr.Markdown("### Nearest Neighbor: Baseline outputs (top 10 prompts)")
-        nn_items = []
-        with gr.Row():
-            for i in range(10):
-                with gr.Column(min_width=120):
-                    nn_items.append(gr.Image(label=f"NN{i+1}", show_label=True, height=140))
-                    nn_items.append(gr.Audio(label="Mix", show_label=True))
-        nn_outputs = nn_items
+        nn_outputs = []
+        for i in range(10):
+            with gr.Row():
+                nn_p_md = gr.Markdown(value="")
+                nn_outputs.append(nn_p_md)
+                nn_img = gr.Image(label=f"NN{i+1}", show_label=True, height=160)
+                nn_m = gr.Audio(label="Mix", show_label=True)
+                nn_outputs.extend([nn_img, nn_m])
 
     # ---- Results View: 3 prompts × 4 methods ----
     with gr.Column(visible=False) as res_col: