v2

gradio-web/gradio_callbacks.py

@@ -383,6 +383,7 @@ def switch_to_execute_phase(uid):
         gr.update(interactive=False),  # next_task_btn
         _live_obs_update(interactive=False),  # img_display
         gr.update(interactive=False),  # reference_action_btn
+        gr.update(interactive=False),  # task_hint_display
     )
 
 
@@ -400,26 +401,50 @@ def switch_to_action_phase(uid=None):
     )
 
 
-def on_video_end_transition(uid, ui_phase=None):
-    """Transition from video phase back to the action phase."""
+def on_demo_video_end_transition(uid, ui_phase=None):
+    """Transition from demo video phase back to the action phase."""
     LOGGER.debug(
-        "on_video_end_transition uid=%s ui_phase=%s",
+        "on_demo_video_end_transition uid=%s ui_phase=%s",
         _uid_for_log(uid),
         ui_phase,
     )
-    log_update = gr.update()
-    if ui_phase == "demo_video" or ui_phase is None:
-        log_update = _action_selection_log()
     return (
         gr.update(visible=False),  # video_phase_group
         gr.update(visible=True),   # action_phase_group
         gr.update(visible=True),   # control_panel_group
-        log_update,  # log_output
+        _action_selection_log(),  # log_output
         gr.update(visible=False, interactive=False),  # watch_demo_video_btn
         "action_point",  # ui_phase_state
     )
 
 
+def on_video_end_transition(uid, ui_phase=None):
+    """Backward-compatible alias for legacy tests and demo end handling."""
+    return on_demo_video_end_transition(uid, ui_phase)
+
+
+def on_execute_video_end_transition(uid, exec_btn_interactive=True):
+    """Transition from execute video phase back to the action phase."""
+    LOGGER.debug(
+        "on_execute_video_end_transition uid=%s exec_btn_interactive=%s",
+        _uid_for_log(uid),
+        exec_btn_interactive,
+    )
+    return (
+        gr.update(visible=False),  # execution_video_group
+        gr.update(visible=True),   # action_phase_group
+        gr.update(visible=True),   # control_panel_group
+        gr.update(interactive=True),  # options_radio
+        gr.update(interactive=bool(exec_btn_interactive)),  # exec_btn
+        gr.update(interactive=True),  # restart_episode_btn
+        gr.update(interactive=True),  # next_task_btn
+        _live_obs_update(interactive=False),  # img_display
+        gr.update(interactive=True),  # reference_action_btn
+        gr.update(interactive=True),  # task_hint_display
+        "action_point",  # ui_phase_state
+    )
+
+
 def _task_load_failed_response(uid, message):
     LOGGER.warning("task_load_failed uid=%s message=%s", _uid_for_log(uid), message)
     _clear_execution_video_path(uid)
@@ -432,6 +457,7 @@ def _task_load_failed_response(uid, message):
         "",  # goal_box
         _ui_text("coords", "not_needed"),  # coords_box
         gr.update(value=None, visible=False, autoplay=False, playback_position=0),  # video_display
+        gr.update(value=None, visible=False, playback_position=0),  # execute_video_display
         gr.update(visible=False, interactive=False),  # watch_demo_video_btn
         "",  # task_info_box
         "",  # progress_info_box
@@ -439,6 +465,7 @@ def _task_load_failed_response(uid, message):
         gr.update(interactive=False),  # next_task_btn
         gr.update(interactive=False),  # exec_btn
         gr.update(visible=False),  # video_phase_group
+        gr.update(visible=False),  # execution_video_group
         gr.update(visible=False),  # action_phase_group
         gr.update(visible=False),  # control_panel_group
         gr.update(value=""),  # task_hint_display
@@ -507,6 +534,7 @@ def _load_status_task(uid, status):
             "",  # goal_box
             _ui_text("coords", "not_needed"),  # coords_box
             gr.update(value=None, visible=False, autoplay=False, playback_position=0),  # video_display
+            gr.update(value=None, visible=False, playback_position=0),  # execute_video_display
             gr.update(visible=False, interactive=False),  # watch_demo_video_btn
             f"{actual_env_id} (Episode {ep_num})",  # task_info_box
             progress_text,  # progress_info_box
@@ -514,6 +542,7 @@ def _load_status_task(uid, status):
             gr.update(interactive=True),  # next_task_btn
             gr.update(interactive=False),  # exec_btn
             gr.update(visible=False),  # video_phase_group
+            gr.update(visible=False),  # execution_video_group
             gr.update(visible=True),  # action_phase_group
             gr.update(visible=True),  # control_panel_group
             gr.update(value=get_task_hint(env_id) if env_id else ""),  # task_hint_display
@@ -574,6 +603,7 @@ def _load_status_task(uid, status):
             goal_text,  # goal_box
             _ui_text("coords", "not_needed"),  # coords_box
             gr.update(value=demo_video_path, visible=True, autoplay=False, playback_position=0),  # video_display
+            gr.update(value=None, visible=False, playback_position=0),  # execute_video_display
             gr.update(visible=True, interactive=True),  # watch_demo_video_btn
             f"{actual_env_id} (Episode {ep_num})",  # task_info_box
             progress_text,  # progress_info_box
@@ -581,6 +611,7 @@ def _load_status_task(uid, status):
             gr.update(interactive=True),  # next_task_btn
             gr.update(interactive=True),  # exec_btn
             gr.update(visible=True),  # video_phase_group
+            gr.update(visible=False),  # execution_video_group
             gr.update(visible=False),  # action_phase_group
             gr.update(visible=False),  # control_panel_group
             gr.update(value=get_task_hint(actual_env_id)),  # task_hint_display
@@ -598,6 +629,7 @@ def _load_status_task(uid, status):
         goal_text,  # goal_box
         _ui_text("coords", "not_needed"),  # coords_box
         gr.update(value=None, visible=False, autoplay=False, playback_position=0),  # video_display (no video)
+        gr.update(value=None, visible=False, playback_position=0),  # execute_video_display
         gr.update(visible=False, interactive=False),  # watch_demo_video_btn
         f"{actual_env_id} (Episode {ep_num})",  # task_info_box
         progress_text,  # progress_info_box
@@ -605,6 +637,7 @@ def _load_status_task(uid, status):
         gr.update(interactive=True),  # next_task_btn
         gr.update(interactive=True),  # exec_btn
         gr.update(visible=False),  # video_phase_group
+        gr.update(visible=False),  # execution_video_group
         gr.update(visible=True),  # action_phase_group
         gr.update(visible=True),  # control_panel_group
         gr.update(value=get_task_hint(actual_env_id)),  # task_hint_display
@@ -1010,15 +1043,17 @@ def execute_step(uid, option_idx, coords_str):
         restart_update=gr.update(interactive=True),
         next_update=gr.update(interactive=True),
         exec_update=gr.update(interactive=True),
-        video_update=None,
+        execute_video_update=None,
         options_update=gr.update(interactive=True),
         coords_update=None,
         reference_update=gr.update(interactive=True),
+        task_hint_update=gr.update(interactive=True),
+        post_execute_exec_interactive=True,
         show_execution_video=False,
         ui_phase="action_point",
     ):
-        if video_update is None:
-            video_update = gr.update(value=None, visible=False, autoplay=False, playback_position=0)
+        if execute_video_update is None:
+            execute_video_update = gr.update(value=None, visible=False, playback_position=0)
         if coords_update is None:
             coords_update = _ui_text("coords", "not_needed")
         return (
@@ -1029,14 +1064,15 @@ def execute_step(uid, option_idx, coords_str):
             restart_update,
             next_update,
             exec_update,
-            video_update,
-            gr.update(visible=False, interactive=False),  # watch_demo_video_btn
-            gr.update(visible=show_execution_video),  # video_phase_group
+            execute_video_update,
             gr.update(visible=not show_execution_video),  # action_phase_group
-            gr.update(visible=not show_execution_video),  # control_panel_group
+            gr.update(visible=True),  # control_panel_group
+            gr.update(visible=show_execution_video),  # execution_video_group
             options_update,
             coords_update,
             reference_update,
+            task_hint_update,
+            bool(post_execute_exec_interactive),
             ui_phase,
         )
 
@@ -1051,6 +1087,8 @@ def execute_step(uid, option_idx, coords_str):
             exec_update=gr.update(interactive=False),
             options_update=gr.update(interactive=False),
             reference_update=gr.update(interactive=False),
+            task_hint_update=gr.update(interactive=False),
+            post_execute_exec_interactive=False,
             show_execution_video=False,
         )
 
@@ -1092,6 +1130,8 @@ def execute_step(uid, option_idx, coords_str):
             exec_update=gr.update(interactive=True),
             options_update=gr.update(choices=_build_radio_choices(session), value=None, interactive=True),
             reference_update=gr.update(interactive=True),
+            task_hint_update=gr.update(interactive=True),
+            post_execute_exec_interactive=True,
             show_execution_video=False,
         )
 
@@ -1115,6 +1155,8 @@ def execute_step(uid, option_idx, coords_str):
                 options_update=gr.update(choices=_build_radio_choices(session), value=None, interactive=True),
                 coords_update=coords_str,
                 reference_update=gr.update(interactive=True),
+                task_hint_update=gr.update(interactive=True),
+                post_execute_exec_interactive=True,
                 show_execution_video=False,
             )
 
@@ -1246,15 +1288,25 @@ def execute_step(uid, option_idx, coords_str):
     
     # 根据视图模式重新获取图片
     img = session.get_pil_image(use_segmented=USE_SEGMENTED_VIEW)
-    video_update = _build_execution_video_update(uid, session)
-    show_execution_video = video_update.get("visible") is True
+    execute_video_update = _build_execution_video_update(uid, session)
+    show_execution_video = execute_video_update.get("visible") is True
     radio_choices = _build_radio_choices(session)
-    restart_episode_update = gr.update(interactive=True)
-    next_task_update = gr.update(interactive=True)
-    exec_btn_update = gr.update(interactive=False) if done else gr.update(interactive=True)
-    options_update = gr.update(choices=radio_choices, value=None, interactive=True)
+    post_execute_exec_interactive = not done
+    restart_episode_update = gr.update(interactive=False) if show_execution_video else gr.update(interactive=True)
+    next_task_update = gr.update(interactive=False) if show_execution_video else gr.update(interactive=True)
+    exec_btn_update = (
+        gr.update(interactive=False)
+        if show_execution_video
+        else gr.update(interactive=post_execute_exec_interactive)
+    )
+    options_update = gr.update(
+        choices=radio_choices,
+        value=None,
+        interactive=not show_execution_video,
+    )
     coords_update = _ui_text("coords", "not_needed")
-    reference_update = gr.update(interactive=True)
+    reference_update = gr.update(interactive=not show_execution_video)
+    task_hint_update = gr.update(interactive=not show_execution_video)
     
     # 格式化日志消息为 HTML 格式（支持颜色显示）
     formatted_status = format_log_markdown(status)
@@ -1276,10 +1328,12 @@ def execute_step(uid, option_idx, coords_str):
         restart_update=restart_episode_update,
         next_update=next_task_update,
         exec_update=exec_btn_update,
-        video_update=video_update,
+        execute_video_update=execute_video_update,
         options_update=options_update,
         coords_update=coords_update,
         reference_update=reference_update,
+        task_hint_update=task_hint_update,
+        post_execute_exec_interactive=post_execute_exec_interactive,
         show_execution_video=show_execution_video,
         ui_phase="execution_video" if show_execution_video else "action_point",
     )

gradio-web/test/test_live_obs_refresh.py

@@ -57,12 +57,14 @@ def test_execute_step_builds_video_from_last_execution_frames(monkeypatch, reloa
     assert [int(frame[0, 0, 0]) for frame in saved_frames] == [11, 22]
     assert suffix.startswith("execute_")
     assert result[7]["visible"] is True
-    assert result[7]["autoplay"] is True
+    assert result[8]["visible"] is False
     assert result[9]["visible"] is True
-    assert result[10]["visible"] is False
-    assert result[11]["visible"] is False
-    assert result[12]["value"] is None
-    assert result[15] == "execution_video"
+    assert result[10]["visible"] is True
+    assert result[11]["value"] is None
+    assert result[11]["interactive"] is False
+    assert result[14]["interactive"] is False
+    assert result[15] is True
+    assert result[16] == "execution_video"
 
 
 def test_execute_step_falls_back_to_single_frame_clip_when_no_new_frames(monkeypatch, reload_module):
@@ -95,7 +97,8 @@ def test_execute_step_falls_back_to_single_frame_clip_when_no_new_frames(monkeyp
     assert len(captured["frames"]) == 1
     assert int(captured["frames"][0][0, 0, 0]) == 33
     assert result[7]["visible"] is True
-    assert result[15] == "execution_video"
+    assert result[10]["visible"] is True
+    assert result[16] == "execution_video"
 
 
 def test_switch_phase_toggles_live_obs_interactive_without_refresh_queue(reload_module):
@@ -103,11 +106,12 @@ def test_switch_phase_toggles_live_obs_interactive_without_refresh_queue(reload_
     callbacks = reload_module("gradio_callbacks")
 
     to_exec = callbacks.switch_to_execute_phase("uid-3")
-    assert len(to_exec) == 6
+    assert len(to_exec) == 7
     assert to_exec[0].get("interactive") is False
     assert to_exec[4].get("interactive") is False
     assert to_exec[4].get("elem_classes") == config.get_live_obs_elem_classes()
     assert to_exec[5].get("interactive") is False
+    assert to_exec[6].get("interactive") is False
 
     to_action = callbacks.switch_to_action_phase()
     assert len(to_action) == 6

gradio-web/test/test_queue_session_limit_e2e.py

@@ -423,6 +423,7 @@ def test_execute_does_not_use_episode_loading_copy(monkeypatch):
             gr.update(interactive=False),
             gr.update(interactive=False),
             gr.update(interactive=False),
+            gr.update(interactive=False),
         )
 
     def fake_execute_step(uid, option_idx, coords_str):
@@ -435,14 +436,15 @@ def test_execute_does_not_use_episode_loading_copy(monkeypatch):
             gr.update(interactive=True),
             gr.update(interactive=True),
             gr.update(interactive=True),
-            gr.update(value=None, visible=False, autoplay=False, playback_position=0),
-            gr.update(visible=False, interactive=False),
-            gr.update(visible=False),
+            gr.update(value=None, visible=False, playback_position=0),
             gr.update(visible=True),
             gr.update(visible=True),
+            gr.update(visible=False),
             gr.update(choices=[("pick", 0)], value=None, interactive=True),
             "No need for coordinates",
             gr.update(interactive=True),
+            gr.update(interactive=True),
+            True,
             "action_point",
         )
 

gradio-web/test/test_ui_native_layout_contract.py

@@ -143,8 +143,10 @@ def test_native_ui_config_contains_phase_machine_and_precheck_chain(reload_modul
             "right_log_col",
             "control_panel_group",
             "video_phase_group",
+            "execution_video_group",
             "action_phase_group",
             "demo_video",
+            "execute_video",
             "watch_demo_video_btn",
             "live_obs",
             "action_radio",
@@ -184,11 +186,18 @@ def test_native_ui_config_contains_phase_machine_and_precheck_chain(reload_modul
             if comp.get("props", {}).get("elem_id") == "demo_video"
         )
         assert demo_video_comp.get("props", {}).get("autoplay") is False
+        execute_video_comp = next(
+            comp
+            for comp in cfg.get("components", [])
+            if comp.get("props", {}).get("elem_id") == "execute_video"
+        )
+        assert execute_video_comp.get("props", {}).get("autoplay") is True
         component_types = [comp.get("type") for comp in cfg.get("components", [])]
         assert "timer" not in component_types
 
         api_names = [dep.get("api_name") for dep in cfg.get("dependencies", [])]
         assert "on_demo_video_play" in api_names
+        assert "on_execute_video_end_transition" in api_names
         assert "precheck_execute_inputs" in api_names
         assert "switch_to_execute_phase" in api_names
         assert "execute_step" in api_names

gradio-web/test/test_ui_phase_machine_runtime_e2e.py

@@ -251,41 +251,47 @@ def _read_phase_visibility(page) -> dict[str, bool | str | None]:
                 return st.display !== 'none' && st.visibility !== 'hidden' && el.getClientRects().length > 0;
             };
             const videoEl = document.querySelector('#demo_video video');
+            const executeVideoEl = document.querySelector('#execute_video video');
             return {
                 videoPhase: visible('video_phase_group'),
                 video: visible('demo_video'),
+                executionVideoPhase: visible('execution_video_group'),
+                executionVideo: visible('execute_video'),
                 watchButton: visible('watch_demo_video_btn'),
                 actionPhase: visible('action_phase_group'),
                 action: visible('live_obs'),
                 controlPhase: visible('control_panel_group'),
                 control: visible('action_radio'),
                 currentSrc: videoEl ? videoEl.currentSrc : null,
+                executeCurrentSrc: executeVideoEl ? executeVideoEl.currentSrc : null,
             };
         }"""
     )
 
 
-def _read_demo_video_controls(page) -> dict[str, bool | None]:
+def _read_demo_video_controls(page, elem_id: str = "demo_video", button_elem_id: str | None = "watch_demo_video_btn") -> dict[str, bool | None]:
     return page.evaluate(
-        """() => {
+        """({ elemId, buttonElemId }) => {
             const visible = (id) => {
+                if (!id) return false;
                 const el = document.getElementById(id);
                 if (!el) return false;
                 const st = getComputedStyle(el);
                 return st.display !== 'none' && st.visibility !== 'hidden' && el.getClientRects().length > 0;
             };
-            const videoEl = document.querySelector('#demo_video video');
-            const button =
-                document.querySelector('#watch_demo_video_btn button') ||
-                document.querySelector('button#watch_demo_video_btn');
+            const videoEl = document.querySelector(`#${elemId} video`);
+            const button = buttonElemId
+                ? (document.querySelector(`#${buttonElemId} button`) || document.querySelector(`button#${buttonElemId}`))
+                : null;
             return {
-                videoVisible: visible('demo_video'),
-                buttonVisible: visible('watch_demo_video_btn'),
+                videoVisible: visible(elemId),
+                buttonVisible: visible(buttonElemId),
                 buttonDisabled: button ? button.disabled : null,
                 autoplay: videoEl ? videoEl.autoplay : null,
                 paused: videoEl ? videoEl.paused : null,
             };
-        }"""
+        }""",
+        {"elemId": elem_id, "buttonElemId": button_elem_id},
     )
 
 
@@ -293,12 +299,12 @@ def _click_demo_video_button(page) -> None:
     page.locator("#watch_demo_video_btn button, button#watch_demo_video_btn").first.click()
 
 
-def _dispatch_video_event(page, event_name: str) -> bool:
+def _dispatch_video_event(page, event_name: str, elem_id: str = "demo_video") -> bool:
     return page.evaluate(
-        """(eventName) => {
+        """({ eventName, elemId }) => {
             const targets = [
-                document.querySelector('#demo_video video'),
-                document.getElementById('demo_video'),
+                document.querySelector(`#${elemId} video`),
+                document.getElementById(elemId),
             ].filter(Boolean);
             if (!targets.length) return false;
             for (const target of targets) {
@@ -306,7 +312,7 @@ def _dispatch_video_event(page, event_name: str) -> bool:
             }
             return true;
         }""",
-        event_name,
+        {"eventName": event_name, "elemId": elem_id},
     )
 
 
@@ -422,6 +428,7 @@ def phase_machine_ui_url():
     with gr.Blocks(title="Native phase machine test") as demo:
         gr.HTML(f"<style>{ui_layout.CSS}</style>")
         phase_state = gr.State("init")
+        post_execute_exec_state = gr.State(True)
 
         with gr.Column(visible=True, elem_id="login_group") as login_group:
             login_btn = gr.Button("Login", elem_id="login_btn")
@@ -436,6 +443,9 @@ def phase_machine_ui_url():
                     visible=False,
                 )
 
+            with gr.Column(visible=False, elem_id="execution_video_group") as execution_video_group:
+                execute_video_display = gr.Video(value=None, elem_id="execute_video", autoplay=True)
+
             with gr.Column(visible=False, elem_id="action_phase_group") as action_phase_group:
                 img_display = gr.Image(value=np.zeros((24, 24, 3), dtype=np.uint8), elem_id="live_obs")
 
@@ -450,6 +460,7 @@ def phase_machine_ui_url():
                         interactive=False,
                     )
                     next_task_btn = gr.Button("Next Task", elem_id="next_task_btn")
+                task_hint_display = gr.Textbox(value="hint", interactive=True, elem_id="task_hint_display")
 
         log_output = gr.Markdown("", elem_id="log_output")
         simulate_stop_btn = gr.Button("Simulate Stop", elem_id="simulate_stop_btn")
@@ -472,6 +483,7 @@ def phase_machine_ui_url():
                 gr.update(visible=False),
                 gr.update(interactive=False),
                 gr.update(value="please click the point selection image"),
+                gr.update(visible=False),
                 "demo_video",
             )
 
@@ -493,6 +505,20 @@ def phase_machine_ui_url():
                 "action_point",
             )
 
+        def on_execute_video_end_fn(exec_enabled):
+            return (
+                gr.update(visible=False),
+                gr.update(visible=True),
+                gr.update(visible=True),
+                gr.update(interactive=True),
+                gr.update(interactive=bool(exec_enabled)),
+                gr.update(interactive=True),
+                gr.update(interactive=False),
+                gr.update(interactive=True),
+                gr.update(interactive=True),
+                "action_point",
+            )
+
         def precheck_fn(_option_idx, _coords):
             state["precheck_calls"] += 1
             if state["precheck_calls"] == 1:
@@ -505,22 +531,23 @@ def phase_machine_ui_url():
                 gr.update(interactive=False),
                 gr.update(interactive=False),
                 gr.update(interactive=False),
+                gr.update(interactive=False),
             )
 
         def execute_fn():
             time.sleep(0.8)
             return (
                 "executed",
-                gr.update(interactive=True),
-                gr.update(interactive=True),
-                gr.update(value=execution_video_path, visible=True, autoplay=True, playback_position=0),
-                gr.update(visible=False, interactive=False),
-                gr.update(visible=True),
-                gr.update(visible=False),
                 gr.update(visible=False),
-                gr.update(interactive=True),
+                gr.update(interactive=False),
+                gr.update(interactive=False),
+                gr.update(value=execution_video_path, visible=True, playback_position=0),
+                gr.update(visible=True),
+                gr.update(interactive=False),
                 "No need for coordinates",
-                gr.update(interactive=True),
+                gr.update(interactive=False),
+                gr.update(interactive=False),
+                True,
                 "execution_video",
             )
 
@@ -537,6 +564,7 @@ def phase_machine_ui_url():
                 action_buttons_row,
                 reference_action_btn,
                 coords_box,
+                execution_video_group,
                 phase_state,
             ],
             queue=False,
@@ -574,6 +602,40 @@ def phase_machine_ui_url():
             ],
             queue=False,
         )
+        execute_video_display.end(
+            fn=on_execute_video_end_fn,
+            inputs=[post_execute_exec_state],
+            outputs=[
+                execution_video_group,
+                action_phase_group,
+                control_panel_group,
+                options_radio,
+                exec_btn,
+                next_task_btn,
+                img_display,
+                reference_action_btn,
+                task_hint_display,
+                phase_state,
+            ],
+            queue=False,
+        )
+        execute_video_display.stop(
+            fn=on_execute_video_end_fn,
+            inputs=[post_execute_exec_state],
+            outputs=[
+                execution_video_group,
+                action_phase_group,
+                control_panel_group,
+                options_radio,
+                exec_btn,
+                next_task_btn,
+                img_display,
+                reference_action_btn,
+                task_hint_display,
+                phase_state,
+            ],
+            queue=False,
+        )
         simulate_stop_btn.click(
             fn=on_simulate_stop_fn,
             outputs=[log_output],
@@ -585,6 +647,8 @@ def phase_machine_ui_url():
                 };
                 show('video_phase_group', false);
                 show('demo_video', false);
+                show('execution_video_group', false);
+                show('execute_video', false);
                 show('action_phase_group', true);
                 show('live_obs', true);
                 show('control_panel_group', true);
@@ -615,22 +679,23 @@ def phase_machine_ui_url():
                 next_task_btn,
                 img_display,
                 reference_action_btn,
+                task_hint_display,
             ],
             queue=False,
         ).then(
             fn=execute_fn,
             outputs=[
                 log_output,
-                next_task_btn,
-                exec_btn,
-                video_display,
-                watch_demo_video_btn,
-                video_phase_group,
                 action_phase_group,
-                control_panel_group,
+                exec_btn,
+                next_task_btn,
+                execute_video_display,
+                execution_video_group,
                 options_radio,
                 coords_box,
                 reference_action_btn,
+                task_hint_display,
+                post_execute_exec_state,
                 phase_state,
             ],
             queue=False,
@@ -972,17 +1037,37 @@ def test_phase_machine_runtime_flow_and_execute_precheck(phase_machine_ui_url):
 
         page.wait_for_function(
             """() => {
-                const videoEl = document.querySelector('#demo_video video');
+                const videoEl = document.querySelector('#execute_video video');
                 return !!videoEl && videoEl.autoplay === true && (videoEl.paused === false || videoEl.currentTime > 0);
             }""",
             timeout=6000,
         )
-        execute_video_controls = _read_demo_video_controls(page)
-        assert execute_video_controls["videoVisible"] is True
+        execute_video_controls = _read_demo_video_controls(page, elem_id="execute_video", button_elem_id=None)
         assert execute_video_controls["autoplay"] is True
         assert execute_video_controls["paused"] is False
+        execute_phase_snapshot = _read_phase_visibility(page)
+        assert execute_phase_snapshot["actionPhase"] is False
+        assert execute_phase_snapshot["controlPhase"] is True
+        panel_snapshot = page.evaluate(
+            """() => {
+                const resolveButton = (id) => {
+                    return document.querySelector(`#${id} button`) || document.querySelector(`button#${id}`);
+                };
+                const radio = document.querySelector('#action_radio input[type="radio"]');
+                const refBtn = resolveButton('reference_action_btn');
+                const hint = document.querySelector('#task_hint_display textarea, #task_hint_display input');
+                return {
+                    radioDisabled: radio ? radio.disabled : null,
+                    refDisabled: refBtn ? refBtn.disabled : null,
+                    hintDisabled: hint ? hint.disabled : null,
+                };
+            }"""
+        )
+        assert panel_snapshot["radioDisabled"] is True
+        assert panel_snapshot["refDisabled"] is True
+        assert panel_snapshot["hintDisabled"] is True
 
-        did_dispatch_end = _dispatch_video_event(page, "ended")
+        did_dispatch_end = _dispatch_video_event(page, "ended", elem_id="execute_video")
         assert did_dispatch_end
 
         page.wait_for_function(
@@ -2523,6 +2608,7 @@ def test_phase_machine_runtime_local_video_path_end_transition():
         with gr.Blocks(title="Native phase machine local video test") as demo:
             uid_state = gr.State(value="uid-local-video")
             phase_state = gr.State(value="action_point")
+            post_execute_exec_state = gr.State(value=True)
             suppress_state = gr.State(value=False)
             with gr.Column(visible=True, elem_id="main_interface") as main_interface:
                 with gr.Column(visible=False, elem_id="video_phase_group") as video_phase_group:
@@ -2534,6 +2620,9 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                         visible=False,
                     )
 
+                with gr.Column(visible=False, elem_id="execution_video_group") as execution_video_group:
+                    execute_video_display = gr.Video(value=None, elem_id="execute_video", autoplay=True)
+
                 with gr.Column(visible=True, elem_id="action_phase_group") as action_phase_group:
                     img_display = gr.Image(value=fake_obs.copy(), elem_id="live_obs")
 
@@ -2544,6 +2633,7 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                     reference_action_btn = gr.Button("reference", interactive=True, elem_id="reference_action_btn")
                     restart_episode_btn = gr.Button("restart", interactive=True, elem_id="restart_episode_btn")
                     next_task_btn = gr.Button("next", interactive=True, elem_id="next_task_btn")
+                    task_hint_display = gr.Textbox("hint", interactive=True, elem_id="task_hint_display")
 
             log_output = gr.Markdown("", elem_id="log_output")
             task_info_box = gr.Textbox("")
@@ -2564,6 +2654,7 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                     next_task_btn,
                     img_display,
                     reference_action_btn,
+                    task_hint_display,
                 ],
                 queue=False,
             ).then(
@@ -2577,14 +2668,15 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                     restart_episode_btn,
                     next_task_btn,
                     exec_btn,
-                    video_display,
-                    watch_demo_video_btn,
-                    video_phase_group,
+                    execute_video_display,
                     action_phase_group,
                     control_panel_group,
+                    execution_video_group,
                     options_radio,
                     coords_box,
                     reference_action_btn,
+                    task_hint_display,
+                    post_execute_exec_state,
                     phase_state,
                 ],
                 queue=False,
@@ -2596,28 +2688,38 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                 queue=False,
             )
 
-            video_display.end(
-                fn=cb.on_video_end_transition,
-                inputs=[uid_state, phase_state],
+            execute_video_display.end(
+                fn=cb.on_execute_video_end_transition,
+                inputs=[uid_state, post_execute_exec_state],
                 outputs=[
-                    video_phase_group,
+                    execution_video_group,
                     action_phase_group,
                     control_panel_group,
-                    log_output,
-                    watch_demo_video_btn,
+                    options_radio,
+                    exec_btn,
+                    restart_episode_btn,
+                    next_task_btn,
+                    img_display,
+                    reference_action_btn,
+                    task_hint_display,
                     phase_state,
                 ],
                 queue=False,
             )
-            video_display.stop(
-                fn=cb.on_video_end_transition,
-                inputs=[uid_state, phase_state],
+            execute_video_display.stop(
+                fn=cb.on_execute_video_end_transition,
+                inputs=[uid_state, post_execute_exec_state],
                 outputs=[
-                    video_phase_group,
+                    execution_video_group,
                     action_phase_group,
                     control_panel_group,
-                    log_output,
-                    watch_demo_video_btn,
+                    options_radio,
+                    exec_btn,
+                    restart_episode_btn,
+                    next_task_btn,
+                    img_display,
+                    reference_action_btn,
+                    task_hint_display,
                     phase_state,
                 ],
                 queue=False,
@@ -2644,7 +2746,7 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                 page.wait_for_selector("#main_interface", state="visible", timeout=20000)
                 page.locator("#action_radio input[type='radio']").first.check(force=True)
                 page.locator("#exec_btn button, button#exec_btn").first.click()
-                page.wait_for_selector("#demo_video video", timeout=5000)
+                page.wait_for_selector("#execute_video video", timeout=5000)
                 page.wait_for_function(
                     """() => {
                         const visible = (id) => {
@@ -2653,13 +2755,12 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                             const st = getComputedStyle(el);
                             return st.display !== 'none' && st.visibility !== 'hidden' && el.getClientRects().length > 0;
                         };
-                        const videoEl = document.querySelector('#demo_video video');
+                        const videoEl = document.querySelector('#execute_video video');
                         return (
-                            visible('video_phase_group') &&
-                            visible('demo_video') &&
-                            !visible('watch_demo_video_btn') &&
+                            visible('execution_video_group') &&
+                            visible('execute_video') &&
                             !visible('action_phase_group') &&
-                            !visible('control_panel_group') &&
+                            visible('control_panel_group') &&
                             !!videoEl &&
                             videoEl.autoplay === true &&
                             (videoEl.paused === false || videoEl.currentTime > 0)
@@ -2667,13 +2768,37 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                     }""",
                     timeout=10000,
                 )
-                controls_after_execute = _read_demo_video_controls(page)
-                assert controls_after_execute["videoVisible"] is True
-                assert controls_after_execute["buttonVisible"] is False
+                controls_after_execute = _read_demo_video_controls(page, elem_id="execute_video", button_elem_id=None)
                 assert controls_after_execute["autoplay"] is True
                 assert controls_after_execute["paused"] is False
+                panel_snapshot = page.evaluate(
+                    """() => {
+                        const resolveButton = (id) => {
+                            return document.querySelector(`#${id} button`) || document.querySelector(`button#${id}`);
+                        };
+                        const radio = document.querySelector('#action_radio input[type="radio"]');
+                        const refBtn = resolveButton('reference_action_btn');
+                        const restartBtn = resolveButton('restart_episode_btn');
+                        const nextBtn = resolveButton('next_task_btn');
+                        const hint = document.querySelector('#task_hint_display textarea, #task_hint_display input');
+                        return {
+                            radioDisabled: radio ? radio.disabled : null,
+                            refDisabled: refBtn ? refBtn.disabled : null,
+                            restartDisabled: restartBtn ? restartBtn.disabled : null,
+                            nextDisabled: nextBtn ? nextBtn.disabled : null,
+                            hintDisabled: hint ? hint.disabled : null,
+                        };
+                    }"""
+                )
+                assert panel_snapshot == {
+                    "radioDisabled": True,
+                    "refDisabled": True,
+                    "restartDisabled": True,
+                    "nextDisabled": True,
+                    "hintDisabled": True,
+                }
 
-                did_dispatch_end = _dispatch_video_event(page, "ended")
+                did_dispatch_end = _dispatch_video_event(page, "ended", elem_id="execute_video")
                 assert did_dispatch_end
 
                 page.wait_for_function(
@@ -2687,8 +2812,8 @@ def test_phase_machine_runtime_local_video_path_end_transition():
                         return (
                             visible('live_obs') &&
                             visible('action_radio') &&
-                            !visible('demo_video') &&
-                            !visible('watch_demo_video_btn')
+                            !visible('execute_video') &&
+                            visible('control_panel_group')
                         );
                     }""",
                     timeout=2000,

gradio-web/test/test_ui_text_config.py

@@ -83,6 +83,23 @@ def test_on_video_end_transition_uses_configured_action_prompt(monkeypatch, relo
     assert result[5] == "action_point"
 
 
+def test_on_execute_video_end_transition_preserves_log_and_restores_controls(reload_module):
+    callbacks = reload_module("gradio_callbacks")
+
+    result = callbacks.on_execute_video_end_transition("uid-1", False)
+
+    assert result[0]["visible"] is False
+    assert result[1]["visible"] is True
+    assert result[2]["visible"] is True
+    assert result[3]["interactive"] is True
+    assert result[4]["interactive"] is False
+    assert result[5]["interactive"] is True
+    assert result[6]["interactive"] is True
+    assert result[8]["interactive"] is True
+    assert result[9]["interactive"] is True
+    assert result[10] == "action_point"
+
+
 def test_on_demo_video_play_disables_button_and_sets_single_use_state(monkeypatch, reload_module):
     reload_module("config")
     callbacks = reload_module("gradio_callbacks")
@@ -234,11 +251,12 @@ def test_load_status_task_shows_demo_video_button_for_valid_video(monkeypatch, r
 
     assert result[7]["visible"] is True
     assert result[7]["value"] == str(video_path)
-    assert result[8]["visible"] is True
-    assert result[8]["interactive"] is True
-    assert result[14]["visible"] is True
-    assert result[15]["visible"] is False
+    assert result[8]["visible"] is False
+    assert result[9]["visible"] is True
+    assert result[9]["interactive"] is True
+    assert result[15]["visible"] is True
     assert result[16]["visible"] is False
+    assert result[17]["visible"] is False
     assert callbacks.UI_TEXT["log"]["demo_video_prompt"] in result[3]
 
 
@@ -268,10 +286,12 @@ def test_load_status_task_hides_demo_video_button_when_video_is_missing(monkeypa
 
     assert result[7]["visible"] is False
     assert result[8]["visible"] is False
-    assert result[8]["interactive"] is False
-    assert result[14]["visible"] is False
-    assert result[15]["visible"] is True
-    assert result[16]["visible"] is True
+    assert result[9]["visible"] is False
+    assert result[9]["interactive"] is False
+    assert result[15]["visible"] is False
+    assert result[16]["visible"] is False
+    assert result[17]["visible"] is True
+    assert result[18]["visible"] is True
     assert callbacks.UI_TEXT["log"]["action_selection_prompt"] in result[3]
 
 

gradio-web/ui_layout.py

@@ -31,13 +31,13 @@ from gradio_callbacks import (
     load_next_task_wrapper,
     on_map_click,
     on_demo_video_play,
+    on_demo_video_end_transition,
+    on_execute_video_end_transition,
     on_option_select,
     on_reference_action,
-    on_video_end_transition,
     precheck_execute_inputs,
     restart_episode_wrapper,
     switch_env_wrapper,
-    switch_to_action_phase,
     switch_to_execute_phase,
     touch_session,
 )
@@ -740,12 +740,15 @@ button#watch_demo_video_btn {{
 #media_card > div,
 #media_card #action_phase_group,
 #media_card #video_phase_group,
+#media_card #execution_video_group,
 #media_card #live_obs,
 #media_card #live_obs button,
 #media_card #live_obs .image-frame,
 #media_card #live_obs img,
 #media_card #demo_video,
-#media_card #demo_video video {{
+#media_card #demo_video video,
+#media_card #execute_video,
+#media_card #execute_video video {{
     border-radius: var(--media-card-radius);
 }}
 
@@ -835,16 +838,17 @@ def render_header_goal(goal_text):
     return capitalize_first_letter(last_goal)
 
 
-def _phase_from_updates(main_interface_update, video_phase_update):
+def _phase_from_updates(main_interface_update, demo_video_phase_update):
     if isinstance(main_interface_update, dict) and main_interface_update.get("visible") is False:
         return PHASE_INIT
-    if isinstance(video_phase_update, dict) and video_phase_update.get("visible") is True:
+    if isinstance(demo_video_phase_update, dict) and demo_video_phase_update.get("visible") is True:
         return PHASE_DEMO_VIDEO
     return PHASE_ACTION_POINT
 
 
 def _with_phase_from_load(load_result):
-    phase = _phase_from_updates(load_result[1], load_result[14])
+    load_result = _normalize_load_result(load_result)
+    phase = _phase_from_updates(load_result[1], load_result[15])
     return (
         *load_result,
         phase,
@@ -853,6 +857,7 @@ def _with_phase_from_load(load_result):
 
 
 def _with_rejected_init(load_result, message):
+    load_result = _normalize_load_result(load_result)
     return (
         *load_result,
         PHASE_INIT,
@@ -860,15 +865,34 @@ def _with_rejected_init(load_result, message):
     )
 
 
+def _normalize_load_result(load_result):
+    if not isinstance(load_result, (tuple, list)):
+        return load_result
+    normalized = list(load_result)
+    if len(normalized) in {19, 20}:
+        normalized.insert(8, gr.update(value=None, visible=False, playback_position=0))
+        normalized.insert(16, gr.update(visible=False))
+    return tuple(normalized)
+
+
 def _phase_visibility_updates(phase):
-    if phase in {PHASE_DEMO_VIDEO, PHASE_EXECUTION_VIDEO}:
+    if phase == PHASE_DEMO_VIDEO:
         return (
             gr.update(visible=True),
             gr.update(visible=False),
             gr.update(visible=False),
+            gr.update(visible=False),
+        )
+    if phase == PHASE_EXECUTION_VIDEO:
+        return (
+            gr.update(visible=False),
+            gr.update(visible=True),
+            gr.update(visible=False),
+            gr.update(visible=True),
         )
     if phase == PHASE_ACTION_POINT:
         return (
+            gr.update(visible=False),
             gr.update(visible=False),
             gr.update(visible=True),
             gr.update(visible=True),
@@ -877,6 +901,7 @@ def _phase_visibility_updates(phase):
         gr.update(visible=False),
         gr.update(visible=False),
         gr.update(visible=False),
+        gr.update(visible=False),
     )
 
 
@@ -927,6 +952,7 @@ def create_ui_blocks():
             delete_callback=cleanup_user_session,
         )
         ui_phase_state = gr.State(value=PHASE_INIT)
+        post_execute_exec_interactive_state = gr.State(value=True)
         current_task_env_state = gr.State(value=None)
         suppress_next_option_change_state = gr.State(value=False)
 
@@ -962,6 +988,16 @@ def create_ui_blocks():
                                 elem_id="watch_demo_video_btn",
                             )
 
+                        with gr.Column(visible=False, elem_id="execution_video_group") as execution_video_group:
+                            execute_video_display = gr.Video(
+                                label="Execution Playback 🎬",
+                                interactive=False,
+                                elem_id="execute_video",
+                                autoplay=True,
+                                show_label=True,
+                                visible=True,
+                            )
+
                         with gr.Column(visible=False, elem_id="action_phase_group") as action_phase_group:
                             img_display = gr.Image(
                                 label="Point Selection",
@@ -1069,6 +1105,7 @@ def create_ui_blocks():
             goal_box,
             coords_box,
             video_display,
+            execute_video_display,
             watch_demo_video_btn,
             task_info_box,
             progress_info_box,
@@ -1076,6 +1113,7 @@ def create_ui_blocks():
             next_task_btn,
             exec_btn,
             video_phase_group,
+            execution_video_group,
             action_phase_group,
             control_panel_group,
             task_hint_display,
@@ -1085,6 +1123,7 @@ def create_ui_blocks():
         ]
         phase_visibility_outputs = [
             video_phase_group,
+            execution_video_group,
             action_phase_group,
             control_panel_group,
         ]
@@ -1312,7 +1351,7 @@ def create_ui_blocks():
         )
 
         video_display.end(
-            fn=on_video_end_transition,
+            fn=on_demo_video_end_transition,
             inputs=[uid_state, ui_phase_state],
             outputs=[
                 video_phase_group,
@@ -1332,7 +1371,7 @@ def create_ui_blocks():
             show_progress="hidden",
         )
         video_display.stop(
-            fn=on_video_end_transition,
+            fn=on_demo_video_end_transition,
             inputs=[uid_state, ui_phase_state],
             outputs=[
                 video_phase_group,
@@ -1352,6 +1391,57 @@ def create_ui_blocks():
             show_progress="hidden",
         )
 
+        execute_video_display.end(
+            fn=on_execute_video_end_transition,
+            inputs=[uid_state, post_execute_exec_interactive_state],
+            outputs=[
+                execution_video_group,
+                action_phase_group,
+                control_panel_group,
+                options_radio,
+                exec_btn,
+                restart_episode_btn,
+                next_task_btn,
+                img_display,
+                reference_action_btn,
+                task_hint_display,
+                ui_phase_state,
+            ],
+            queue=False,
+            show_progress="hidden",
+        ).then(
+            fn=touch_session,
+            inputs=[uid_state],
+            outputs=[uid_state],
+            queue=False,
+            show_progress="hidden",
+        )
+        execute_video_display.stop(
+            fn=on_execute_video_end_transition,
+            inputs=[uid_state, post_execute_exec_interactive_state],
+            outputs=[
+                execution_video_group,
+                action_phase_group,
+                control_panel_group,
+                options_radio,
+                exec_btn,
+                restart_episode_btn,
+                next_task_btn,
+                img_display,
+                reference_action_btn,
+                task_hint_display,
+                ui_phase_state,
+            ],
+            queue=False,
+            show_progress="hidden",
+        ).then(
+            fn=touch_session,
+            inputs=[uid_state],
+            outputs=[uid_state],
+            queue=False,
+            show_progress="hidden",
+        )
+
         img_display.select(
             fn=on_map_click,
             inputs=[uid_state, options_radio],
@@ -1423,6 +1513,7 @@ def create_ui_blocks():
                 next_task_btn,
                 img_display,
                 reference_action_btn,
+                task_hint_display,
             ],
             queue=False,
             show_progress="hidden",
@@ -1443,14 +1534,15 @@ def create_ui_blocks():
                 restart_episode_btn,
                 next_task_btn,
                 exec_btn,
-                video_display,
-                watch_demo_video_btn,
-                video_phase_group,
+                execute_video_display,
                 action_phase_group,
                 control_panel_group,
+                execution_video_group,
                 options_radio,
                 coords_box,
                 reference_action_btn,
+                task_hint_display,
+                post_execute_exec_interactive_state,
                 ui_phase_state,
             ],
             show_progress="hidden",