ground truth action only press once!

gradio-web/gradio_callbacks.py

@@ -855,17 +855,25 @@ def _is_valid_coords_text(coords_text: str) -> bool:
     return True
 
 
-def on_option_select(uid, option_value, coords_str=None):
+def on_option_select(uid, option_value, coords_str=None, suppress_next_option_change=False):
     """
     处理选项选择事件
     """
     default_msg = _ui_text("coords", "not_needed")
+
+    if suppress_next_option_change:
+        LOGGER.debug(
+            "on_option_select suppressed uid=%s option=%s",
+            _uid_for_log(uid),
+            option_value,
+        )
+        return gr.update(), gr.update(), gr.update(), False
     
     if option_value is None:
         LOGGER.debug("on_option_select uid=%s option=None", _uid_for_log(uid))
         session = get_session(uid) if uid else None
         base_img = session.get_pil_image(use_segmented=USE_SEGMENTED_VIEW) if session else _LIVE_OBS_UPDATE_SKIP
-        return default_msg, _live_obs_update(value=base_img, interactive=False), _action_selection_log()
+        return default_msg, _live_obs_update(value=base_img, interactive=False), _action_selection_log(), False
     
     # 更新session活动时间（选择选项操作）
     if uid:
@@ -874,7 +882,7 @@ def on_option_select(uid, option_value, coords_str=None):
     session = get_session(uid)
     if not session:
         LOGGER.warning("on_option_select: missing session uid=%s", _uid_for_log(uid))
-        return default_msg, _live_obs_update(interactive=False), format_log_markdown(_ui_text("log", "session_error"))
+        return default_msg, _live_obs_update(interactive=False), format_log_markdown(_ui_text("log", "session_error")), False
     
     option_idx = _parse_option_idx(option_value)
     base_img = session.get_pil_image(use_segmented=USE_SEGMENTED_VIEW)
@@ -891,13 +899,14 @@ def on_option_select(uid, option_value, coords_str=None):
             _ui_text("coords", "select_keypoint"),
             _live_obs_update(value=base_img, interactive=True, waiting_for_keypoint=True),
             _keypoint_selection_log(),
+            False,
         )
     
     LOGGER.debug("on_option_select uid=%s option=%s requires_coords=False", _uid_for_log(uid), option_idx)
-    return default_msg, _live_obs_update(value=base_img, interactive=False), _action_selection_log()
+    return default_msg, _live_obs_update(value=base_img, interactive=False), _action_selection_log(), False
 
 
-def on_reference_action(uid):
+def on_reference_action(uid, current_option_value=None):
     """
     自动获取并回填当前步参考 action + 像素坐标（不执行）。
     """
@@ -912,6 +921,7 @@ def on_reference_action(uid):
             gr.update(),
             _ui_text("coords", "not_needed"),
             format_log_markdown(_ui_text("log", "session_error")),
+            False,
         )
 
     LOGGER.info("on_reference_action uid=%s env=%s", _uid_for_log(uid), getattr(session, "env_id", None))
@@ -926,6 +936,7 @@ def on_reference_action(uid):
             gr.update(),
             gr.update(),
             format_log_markdown(_ui_text("log", "reference_action_error", error=exc)),
+            False,
         )
 
     if not isinstance(reference, dict) or not reference.get("ok", False):
@@ -937,14 +948,17 @@ def on_reference_action(uid):
             gr.update(),
             gr.update(),
             format_log_markdown(_ui_text("log", "reference_action_status", message=message)),
+            False,
         )
 
     option_idx = reference.get("option_idx")
+    current_option_idx = _parse_option_idx(current_option_value)
     option_label = str(reference.get("option_label", "")).strip()
     option_action = str(reference.get("option_action", "")).strip()
     option_action = get_ui_action_text(getattr(session, "env_id", None), option_action)
     need_coords = bool(reference.get("need_coords", False))
     coords_xy = reference.get("coords_xy")
+    suppress_next_option_change = option_idx != current_option_idx
 
     updated_img = current_img
     coords_text = _ui_text("coords", "not_needed")
@@ -980,6 +994,7 @@ def on_reference_action(uid):
         gr.update(value=option_idx),
         coords_text,
         format_log_markdown(log_text),
+        suppress_next_option_change,
     )
 
 

gradio-web/test/test_reference_action_callbacks.py

@@ -48,7 +48,7 @@ def test_on_reference_action_success_updates_option_and_coords(monkeypatch, relo
     monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
     monkeypatch.setattr(callbacks, "get_session", lambda uid: session)
 
-    img_update, option_update, coords_text, log_html = callbacks.on_reference_action("uid-1")
+    img_update, option_update, coords_text, log_html, suppress_flag = callbacks.on_reference_action("uid-1", None)
 
     assert img_update.get("__type__") == "update"
     assert isinstance(img_update.get("value"), Image.Image)
@@ -56,6 +56,7 @@ def test_on_reference_action_success_updates_option_and_coords(monkeypatch, relo
     assert img_update.get("elem_classes") == config.get_live_obs_elem_classes()
     assert option_update.get("value") == 2
     assert coords_text == "5, 6"
+    assert suppress_flag is True
     expected_log = config.UI_TEXT["log"]["reference_action_message_with_coords"].format(
         option_label="c",
         option_action="press the button",
@@ -71,13 +72,14 @@ def test_on_reference_action_session_missing(monkeypatch, reload_module):
     monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
     monkeypatch.setattr(callbacks, "get_session", lambda uid: None)
 
-    img_update, option_update, coords_text, log_html = callbacks.on_reference_action("uid-missing")
+    img_update, option_update, coords_text, log_html, suppress_flag = callbacks.on_reference_action("uid-missing", None)
 
     assert img_update.get("__type__") == "update"
     assert img_update.get("value") is None
     assert option_update.get("__type__") == "update"
     assert coords_text == config.UI_TEXT["coords"]["not_needed"]
     assert log_html == config.UI_TEXT["log"]["session_error"]
+    assert suppress_flag is False
 
 
 def test_on_reference_action_error_message_from_reference(monkeypatch, reload_module):
@@ -88,8 +90,33 @@ def test_on_reference_action_error_message_from_reference(monkeypatch, reload_mo
     monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
     monkeypatch.setattr(callbacks, "get_session", lambda uid: session)
 
-    _img, _opt, _coords, log_html = callbacks.on_reference_action("uid-1")
+    _img, _opt, _coords, log_html, suppress_flag = callbacks.on_reference_action("uid-1", None)
     assert log_html == config.UI_TEXT["log"]["reference_action_status"].format(message="bad ref")
+    assert suppress_flag is False
+
+
+def test_on_reference_action_same_selected_option_does_not_set_suppression(monkeypatch, reload_module):
+    callbacks = reload_module("gradio_callbacks")
+
+    session = _FakeSession(
+        {
+            "ok": True,
+            "option_idx": 0,
+            "option_label": "a",
+            "option_action": "pick up the cube",
+            "need_coords": True,
+            "coords_xy": [3, 4],
+            "message": "ok",
+        }
+    )
+
+    monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
+    monkeypatch.setattr(callbacks, "get_session", lambda uid: session)
+
+    _img, _option_update, coords_text, _log_html, suppress_flag = callbacks.on_reference_action("uid-1", 0)
+
+    assert coords_text == "3, 4"
+    assert suppress_flag is False
 
 
 def test_on_option_select_resets_to_keypoint_wait_state_for_point_action(monkeypatch, reload_module):
@@ -100,12 +127,24 @@ def test_on_option_select_resets_to_keypoint_wait_state_for_point_action(monkeyp
     monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
     monkeypatch.setattr(callbacks, "get_session", lambda uid: session)
 
-    coords_text, img_update, log_text = callbacks.on_option_select("uid-1", 0, "12, 34")
+    coords_text, img_update, log_text, suppress_flag = callbacks.on_option_select("uid-1", 0, "12, 34", False)
 
     assert coords_text == config.UI_TEXT["coords"]["select_keypoint"]
     assert img_update.get("interactive") is True
     assert img_update.get("elem_classes") == config.get_live_obs_elem_classes(waiting_for_keypoint=True)
     assert log_text == config.UI_TEXT["log"]["keypoint_selection_prompt"]
+    assert suppress_flag is False
+
+
+def test_on_option_select_suppresses_programmatic_reference_change(reload_module):
+    callbacks = reload_module("gradio_callbacks")
+
+    coords_update, img_update, log_update, suppress_flag = callbacks.on_option_select("uid-1", 0, "12, 34", True)
+
+    assert coords_update.get("__type__") == "update"
+    assert img_update.get("__type__") == "update"
+    assert log_update.get("__type__") == "update"
+    assert suppress_flag is False
 
 
 def test_on_map_click_clears_wait_state_and_restores_action_prompt(monkeypatch, reload_module):
@@ -147,10 +186,11 @@ def test_on_reference_action_uses_configured_action_text_override(monkeypatch, r
     monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
     monkeypatch.setattr(callbacks, "get_session", lambda uid: session)
 
-    _img, _option_update, coords_text, log_html = callbacks.on_reference_action("uid-1")
+    _img, _option_update, coords_text, log_html, suppress_flag = callbacks.on_reference_action("uid-1", None)
 
     assert coords_text == config.UI_TEXT["coords"]["not_needed"]
     assert log_html == config.UI_TEXT["log"]["reference_action_message"].format(
         option_label="a",
         option_action="move forward↓",
     )
+    assert suppress_flag is True

gradio-web/test/test_ui_phase_machine_runtime_e2e.py

@@ -1415,6 +1415,190 @@ def test_keypoint_wait_state_pulses_live_obs_and_updates_system_log(monkeypatch)
         demo.close()
 
 
+def test_reference_action_single_click_applies_coords_without_wait_state(monkeypatch):
+    config_module = importlib.reload(importlib.import_module("config"))
+    callbacks = importlib.reload(importlib.import_module("gradio_callbacks"))
+    ui_layout = importlib.reload(importlib.import_module("ui_layout"))
+
+    fake_obs = np.zeros((24, 48, 3), dtype=np.uint8)
+    fake_obs[:, :] = [15, 20, 25]
+    fake_obs_img = Image.fromarray(fake_obs)
+
+    class FakeSession:
+        env_id = "BinFill"
+        raw_solve_options = [
+            {"label": "a", "action": "pick the left cube", "available": [object()]},
+            {"label": "b", "action": "pick the right cube", "available": [object()]},
+        ]
+        available_options = [
+            ("a. pick the left cube", 0),
+            ("b. pick the right cube", 1),
+        ]
+
+        def get_pil_image(self, use_segmented=False):
+            _ = use_segmented
+            return fake_obs_img.copy()
+
+        def get_reference_action(self):
+            return {
+                "ok": True,
+                "option_idx": 0,
+                "option_label": "a",
+                "option_action": "pick the left cube",
+                "need_coords": True,
+                "coords_xy": [5, 6],
+                "message": "ok",
+            }
+
+    def fake_init_app(_request=None):
+        return (
+            "uid-reference-action",
+            gr.update(visible=True),  # main_interface
+            gr.update(
+                value=fake_obs_img.copy(),
+                interactive=False,
+                elem_classes=config_module.get_live_obs_elem_classes(),
+            ),  # img_display
+            config_module.UI_TEXT["log"]["action_selection_prompt"],  # log_output
+            gr.update(
+                choices=[
+                    ("a. pick the left cube", 0),
+                    ("b. pick the right cube", 1),
+                ],
+                value=None,
+            ),  # options_radio
+            "goal",  # goal_box
+            gr.update(
+                value=config_module.UI_TEXT["coords"]["not_needed"],
+                visible=True,
+                interactive=False,
+            ),  # coords_box
+            gr.update(value=None, visible=False),  # video_display
+            gr.update(visible=False, interactive=False),  # watch_demo_video_btn
+            "BinFill (Episode 1)",  # task_info_box
+            "Completed: 0",  # progress_info_box
+            gr.update(interactive=True),  # restart_episode_btn
+            gr.update(interactive=True),  # next_task_btn
+            gr.update(interactive=True),  # exec_btn
+            gr.update(visible=False),  # video_phase_group
+            gr.update(visible=True),  # action_phase_group
+            gr.update(visible=True),  # control_panel_group
+            gr.update(value="hint"),  # task_hint_display
+            gr.update(visible=False),  # loading_overlay
+            gr.update(interactive=True),  # reference_action_btn
+        )
+
+    monkeypatch.setattr(ui_layout, "init_app", fake_init_app)
+    monkeypatch.setattr(callbacks, "get_session", lambda uid: FakeSession())
+    monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
+
+    demo = ui_layout.create_ui_blocks()
+
+    port = _free_port()
+    host = "127.0.0.1"
+    root_url = f"http://{host}:{port}/"
+
+    app = FastAPI(title="reference-action-single-click-test")
+    app = gr.mount_gradio_app(app, demo, path="/")
+
+    config = uvicorn.Config(app, host=host, port=port, log_level="error")
+    server = uvicorn.Server(config)
+    thread = threading.Thread(target=server.run, daemon=True)
+    thread.start()
+    _wait_http_ready(root_url)
+
+    try:
+        with sync_playwright() as p:
+            browser = p.chromium.launch(headless=True)
+            page = browser.new_page(viewport={"width": 1280, "height": 900})
+            page.goto(root_url, wait_until="domcontentloaded")
+            page.wait_for_selector("#main_interface_root", state="visible", timeout=15000)
+            page.wait_for_selector("#live_obs img", timeout=15000)
+            page.wait_for_selector("#reference_action_btn button, button#reference_action_btn", timeout=15000)
+
+            expected_reference_log = config_module.UI_TEXT["log"]["reference_action_message_with_coords"].format(
+                option_label="a",
+                option_action="pick the left cube",
+                coords_text="5, 6",
+            )
+            page.locator("#reference_action_btn button, button#reference_action_btn").first.click()
+
+            page.wait_for_function(
+                """(state) => {
+                    const coordsRoot = document.getElementById('coords_box');
+                    const coordsField = coordsRoot?.querySelector('textarea, input');
+                    const logRoot = document.getElementById('log_output');
+                    const logField = logRoot?.querySelector('textarea, input');
+                    const liveObs = document.getElementById('live_obs');
+                    const checked = document.querySelector('#action_radio input[type="radio"]:checked');
+                    const coordsValue = coordsField ? coordsField.value.trim() : '';
+                    const logValue = logField ? logField.value.trim() : (logRoot?.textContent || '').trim();
+                    return (
+                        !!checked &&
+                        checked.value === state.checkedValue &&
+                        coordsValue === state.coordsValue &&
+                        logValue === state.logValue &&
+                        !!liveObs &&
+                        !liveObs.classList.contains(state.waitClass)
+                    );
+                }""",
+                arg={
+                    "checkedValue": "0",
+                    "coordsValue": "5, 6",
+                    "logValue": expected_reference_log,
+                    "waitClass": config_module.LIVE_OBS_KEYPOINT_WAIT_CLASS,
+                },
+                timeout=5000,
+            )
+
+            classes_after_reference = _read_elem_classes(page, "live_obs")
+            assert classes_after_reference is not None
+            assert config_module.LIVE_OBS_KEYPOINT_WAIT_CLASS not in classes_after_reference
+            assert _read_coords_box_value(page) == "5, 6"
+            assert _read_log_output_value(page) == expected_reference_log
+
+            page.locator("#action_radio input[type='radio']").nth(1).check(force=True)
+            page.wait_for_function(
+                """(state) => {
+                    const coordsRoot = document.getElementById('coords_box');
+                    const coordsField = coordsRoot?.querySelector('textarea, input');
+                    const logRoot = document.getElementById('log_output');
+                    const logField = logRoot?.querySelector('textarea, input');
+                    const liveObs = document.getElementById('live_obs');
+                    const checked = document.querySelector('#action_radio input[type="radio"]:checked');
+                    const coordsValue = coordsField ? coordsField.value.trim() : '';
+                    const logValue = logField ? logField.value.trim() : (logRoot?.textContent || '').trim();
+                    return (
+                        !!checked &&
+                        checked.value === state.checkedValue &&
+                        coordsValue === state.coordsValue &&
+                        logValue === state.logValue &&
+                        !!liveObs &&
+                        liveObs.classList.contains(state.waitClass)
+                    );
+                }""",
+                arg={
+                    "checkedValue": "1",
+                    "coordsValue": config_module.UI_TEXT["coords"]["select_keypoint"],
+                    "logValue": config_module.UI_TEXT["log"]["keypoint_selection_prompt"],
+                    "waitClass": config_module.LIVE_OBS_KEYPOINT_WAIT_CLASS,
+                },
+                timeout=5000,
+            )
+
+            classes_after_manual_change = _read_elem_classes(page, "live_obs")
+            assert classes_after_manual_change is not None
+            assert config_module.LIVE_OBS_KEYPOINT_WAIT_CLASS in classes_after_manual_change
+            assert _read_coords_box_value(page) == config_module.UI_TEXT["coords"]["select_keypoint"]
+            assert _read_log_output_value(page) == config_module.UI_TEXT["log"]["keypoint_selection_prompt"]
+
+            browser.close()
+    finally:
+        server.should_exit = True
+        thread.join(timeout=10)
+        demo.close()
+
+
 def test_live_obs_client_resize_fills_width_and_keeps_click_mapping(monkeypatch):
     callbacks = importlib.reload(importlib.import_module("gradio_callbacks"))
     ui_layout = importlib.reload(importlib.import_module("ui_layout"))

gradio-web/test/test_ui_text_config.py

@@ -43,12 +43,13 @@ def test_on_option_select_uses_configured_select_keypoint_and_log_messages(monke
     monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
     monkeypatch.setattr(callbacks, "get_session", lambda uid: _FakeOptionSession())
 
-    coords_text, img_update, log_text = callbacks.on_option_select("uid-1", 0, None)
+    coords_text, img_update, log_text, suppress_flag = callbacks.on_option_select("uid-1", 0, None, False)
 
     assert coords_text == "pick a point from config"
     assert img_update.get("interactive") is True
     assert callbacks.get_live_obs_elem_classes(waiting_for_keypoint=True) == img_update.get("elem_classes")
     assert log_text == "custom log prompt from config"
+    assert suppress_flag is False
 
 
 def test_precheck_execute_inputs_uses_configured_before_execute_message(monkeypatch, reload_module):
@@ -112,11 +113,12 @@ def test_missing_session_paths_use_configured_session_error(monkeypatch, reload_
     monkeypatch.setattr(callbacks, "update_session_activity", lambda uid: None)
     monkeypatch.setattr(callbacks, "get_session", lambda uid: None)
 
-    _img, _option_update, coords_text, log_text = callbacks.on_reference_action("uid-missing")
+    _img, _option_update, coords_text, log_text, suppress_flag = callbacks.on_reference_action("uid-missing", None)
     map_img, map_coords, map_log = callbacks.on_map_click("uid-missing", None, None)
 
     assert coords_text == callbacks.UI_TEXT["coords"]["not_needed"]
     assert log_text == "Session Error From Config"
+    assert suppress_flag is False
     assert map_img.get("__type__") == "update"
     assert map_img.get("value") is None
     assert map_coords == callbacks.UI_TEXT["coords"]["not_needed"]

gradio-web/ui_layout.py

@@ -586,6 +586,7 @@ def create_ui_blocks():
         uid_state = gr.State(value=None)
         ui_phase_state = gr.State(value=PHASE_INIT)
         current_task_env_state = gr.State(value=None)
+        suppress_next_option_change_state = gr.State(value=False)
         live_obs_timer = gr.Timer(value=1.0 / LIVE_OBS_REFRESH_HZ, active=True)
 
         task_info_box = gr.Textbox(visible=False, elem_id="task_info_box")
@@ -925,8 +926,8 @@ def create_ui_blocks():
 
         options_radio.change(
             fn=on_option_select,
-            inputs=[uid_state, options_radio, coords_box],
-            outputs=[coords_box, img_display, log_output],
+            inputs=[uid_state, options_radio, coords_box, suppress_next_option_change_state],
+            outputs=[coords_box, img_display, log_output, suppress_next_option_change_state],
         )
 
         watch_demo_video_btn.click(
@@ -939,8 +940,8 @@ def create_ui_blocks():
 
         reference_action_btn.click(
             fn=on_reference_action,
-            inputs=[uid_state],
-            outputs=[img_display, options_radio, coords_box, log_output],
+            inputs=[uid_state, options_radio],
+            outputs=[img_display, options_radio, coords_box, log_output, suppress_next_option_change_state],
         )
 
         exec_btn.click(