Update app.py

app.py

@@ -320,6 +320,12 @@ CUSTOM_CSS = """
     white-space: pre-wrap;
     line-height: 1.6;
 }
+#submit-btn button {
+    min-height: 52px !important;
+    font-size: 1.08rem !important;
+    font-weight: 700 !important;
+    padding: 0.7rem 1.2rem !important;
+}
 """
 
 
@@ -508,7 +514,6 @@ def build_sample_brief_html(sample: Dict[str, Any], index: int, total: int) -> s
 
 def create_app():
     samples = build_pending_samples()
-    sample_map = {s["anon_id"]: s for s in samples}
 
     with gr.Blocks(
         title="VideoEval Movie-Level Evaluation",
@@ -526,6 +531,7 @@ def create_app():
 
         current_idx = gr.State(0)
         evaluator_state = gr.State("anonymous")
+        samples_state = gr.State(samples)
 
         with gr.Row():
             with gr.Column(scale=5, elem_classes=["panel"]):
@@ -550,7 +556,7 @@ def create_app():
                     if not samples else build_sample_brief_html(samples[0], 0, len(samples))
                 )
 
-        gr.Markdown("## 3) Movie-Level 指标评分（1-5）")
+        gr.Markdown("## 3) 评分（1-5）")
         gr.Markdown("<span class='hint'>请先完成 1-5 分评分，未打分无法提交。</span>")
 
         score_widgets: Dict[str, gr.Radio] = {}
@@ -576,32 +582,34 @@ def create_app():
                         score_widgets[key] = gr.Radio(choices=[1, 2, 3, 4, 5], label=f"{key} Score")
 
         final_summary = gr.Textbox(label="Final Summary（可选）", lines=4, placeholder="总结该视频的主要优缺点")
-        submit_btn = gr.Button("提交当前评分并统计", variant="primary")
+        submit_btn = gr.Button("提交", variant="primary", elem_id="submit-btn")
 
-        def _sync_sample_from_dropdown(anon_id: str) -> Tuple[str, str, int]:
+        def _sync_sample_from_dropdown(anon_id: str, curr_samples: List[Dict[str, Any]]) -> Tuple[str, str, int]:
+            sample_map = {s["anon_id"]: s for s in curr_samples}
             if not anon_id or anon_id not in sample_map:
                 return None, "<div class='sample-card'><p class='story-body'>未找到样本</p></div>", 0
-            idx = next(i for i, s in enumerate(samples) if s["anon_id"] == anon_id)
-            sample = samples[idx]
-            return sample["video_path"], build_sample_brief_html(sample, idx, len(samples)), idx
+            idx = next(i for i, s in enumerate(curr_samples) if s["anon_id"] == anon_id)
+            sample = curr_samples[idx]
+            return sample["video_path"], build_sample_brief_html(sample, idx, len(curr_samples)), idx
 
-        def _go_prev(idx: int) -> Tuple[str, str, str, int]:
-            if not samples:
+        def _go_prev(idx: int, curr_samples: List[Dict[str, Any]]) -> Tuple[str, str, str, int]:
+            if not curr_samples:
                 return None, "<div class='sample-card'><p class='story-body'>无可用样本</p></div>", None, 0
             idx = max(0, idx - 1)
-            sample = samples[idx]
-            return sample["video_path"], build_sample_brief_html(sample, idx, len(samples)), sample["anon_id"], idx
+            sample = curr_samples[idx]
+            return sample["video_path"], build_sample_brief_html(sample, idx, len(curr_samples)), sample["anon_id"], idx
 
-        def _go_next(idx: int) -> Tuple[str, str, str, int]:
-            if not samples:
+        def _go_next(idx: int, curr_samples: List[Dict[str, Any]]) -> Tuple[str, str, str, int]:
+            if not curr_samples:
                 return None, "<div class='sample-card'><p class='story-body'>无可用样本</p></div>", None, 0
-            idx = min(len(samples) - 1, idx + 1)
-            sample = samples[idx]
-            return sample["video_path"], build_sample_brief_html(sample, idx, len(samples)), sample["anon_id"], idx
+            idx = min(len(curr_samples) - 1, idx + 1)
+            sample = curr_samples[idx]
+            return sample["video_path"], build_sample_brief_html(sample, idx, len(curr_samples)), sample["anon_id"], idx
 
-        def _submit(evaluator_id: str, anon_id: str, summary: str, *score_vals):
-            if not samples:
+        def _submit(evaluator_id: str, anon_id: str, summary: str, curr_samples: List[Dict[str, Any]], *score_vals):
+            if not curr_samples:
                 return "❌ 没有可提交样本。"
+            sample_map = {s["anon_id"]: s for s in curr_samples}
             if not anon_id or anon_id not in sample_map:
                 return "❌ 请先选择样本。"
             sample = sample_map[anon_id]
@@ -627,19 +635,76 @@ def create_app():
 
             return f"✅ 已保存: `{single_path}`\n\n✅ 已更新方法统计: `{agg_path}`"
 
+        def _refresh_on_load() -> Tuple[Any, Any, str, int, str, List[Dict[str, Any]]]:
+            refreshed_samples = build_pending_samples()
+            if not refreshed_samples:
+                return (
+                    gr.update(choices=[], value=None),
+                    None,
+                    "<div class='sample-card'><p class='story-body'>无可用样本（可能都已评估）</p></div>",
+                    0,
+                    "当前无待评估样本。",
+                    refreshed_samples,
+                )
+
+            first = refreshed_samples[0]
+            return (
+                gr.update(choices=[s["anon_id"] for s in refreshed_samples], value=first["anon_id"]),
+                first["video_path"],
+                build_sample_brief_html(first, 0, len(refreshed_samples)),
+                0,
+                f"已刷新样本池：{len(refreshed_samples)} 个待评估样本。",
+                refreshed_samples,
+            )
+
+        def _refresh_after_submit(submit_msg: str) -> Tuple[Any, Any, str, int, str, List[Dict[str, Any]]]:
+            refreshed_samples = build_pending_samples()
+            submit_msg = (submit_msg or "").strip()
+            if not refreshed_samples:
+                status_msg = f"{submit_msg}\n\n当前无待评估样本。" if submit_msg else "当前无待评估样本。"
+                return (
+                    gr.update(choices=[], value=None),
+                    None,
+                    "<div class='sample-card'><p class='story-body'>无可用样本（可能都已评估）</p></div>",
+                    0,
+                    status_msg,
+                    refreshed_samples,
+                )
+
+            first = refreshed_samples[0]
+            refresh_msg = f"已刷新样本池：{len(refreshed_samples)} 个待评估样本。"
+            status_msg = f"{submit_msg}\n\n{refresh_msg}" if submit_msg else refresh_msg
+            return (
+                gr.update(choices=[s["anon_id"] for s in refreshed_samples], value=first["anon_id"]),
+                first["video_path"],
+                build_sample_brief_html(first, 0, len(refreshed_samples)),
+                0,
+                status_msg,
+                refreshed_samples,
+            )
+
         sample_dropdown.change(
             _sync_sample_from_dropdown,
-            inputs=[sample_dropdown],
+            inputs=[sample_dropdown, samples_state],
             outputs=[movie_video, sample_info, current_idx],
         )
-        prev_btn.click(_go_prev, inputs=[current_idx], outputs=[movie_video, sample_info, sample_dropdown, current_idx])
-        next_btn.click(_go_next, inputs=[current_idx], outputs=[movie_video, sample_info, sample_dropdown, current_idx])
+        prev_btn.click(_go_prev, inputs=[current_idx, samples_state], outputs=[movie_video, sample_info, sample_dropdown, current_idx])
+        next_btn.click(_go_next, inputs=[current_idx, samples_state], outputs=[movie_video, sample_info, sample_dropdown, current_idx])
 
-        submit_inputs = [evaluator_input, sample_dropdown, final_summary]
+        submit_inputs = [evaluator_input, sample_dropdown, final_summary, samples_state]
         for key in BASE_METRIC_KEYS:
             submit_inputs.append(score_widgets[key])
-        submit_btn.click(_submit, inputs=submit_inputs, outputs=[status])
+        submit_evt = submit_btn.click(_submit, inputs=submit_inputs, outputs=[status])
+        submit_evt.then(
+            _refresh_after_submit,
+            inputs=[status],
+            outputs=[sample_dropdown, movie_video, sample_info, current_idx, status, samples_state],
+        )
 
+        app.load(
+            _refresh_on_load,
+            outputs=[sample_dropdown, movie_video, sample_info, current_idx, status, samples_state],
+        )
         app.load(lambda x: x, inputs=[evaluator_input], outputs=[evaluator_state])
 
     return app