feat(security): add security demo

dimos/agents/skills/speak_skill.py

@@ -31,6 +31,8 @@ class SpeakSkill(Module):
     _tts_node: OpenAITTSNode | None = None
     _audio_output: SounddeviceAudioOutput | None = None
     _audio_lock: threading.Lock = threading.Lock()
+    _bg_threads: list[threading.Thread] = []
+    _bg_threads_lock: threading.Lock = threading.Lock()
 
     @rpc
     def start(self) -> None:
@@ -41,6 +43,10 @@ def start(self) -> None:
 
     @rpc
     def stop(self) -> None:
+        with self._bg_threads_lock:
+            threads = list(self._bg_threads)
+        for t in threads:
+            t.join(timeout=10.0)
         if self._tts_node:
             self._tts_node.dispose()
             self._tts_node = None
@@ -50,7 +56,7 @@ def stop(self) -> None:
         super().stop()
 
     @skill
-    def speak(self, text: str) -> str:
+    def speak(self, text: str, blocking: bool = True) -> str:
         """Speak text out loud through the robot's speakers.
 
         USE THIS TOOL AS OFTEN AS NEEDED. People can't normally see what you say in text, but can hear what you speak.
@@ -64,8 +70,32 @@ def speak(self, text: str) -> str:
         if self._tts_node is None:
             return "Error: TTS not initialized"
 
+        if not blocking:
+            thread = threading.Thread(
+                target=self._speak_bg, args=(text,), daemon=True, name="SpeakSkill-bg"
+            )
+            with self._bg_threads_lock:
+                self._bg_threads.append(thread)
+            thread.start()
+            return f"Speaking (non-blocking): {text}"
+
+        return self._speak_blocking(text)
+
+    def _speak_bg(self, text: str) -> None:
+        try:
+            self._speak_blocking(text)
+        finally:
+            with self._bg_threads_lock:
+                self._bg_threads = [
+                    t for t in self._bg_threads if t is not threading.current_thread()
+                ]
+
+    def _speak_blocking(self, text: str) -> str:
         # Use lock to prevent simultaneous speech
         with self._audio_lock:
+            if self._tts_node is None:
+                return "Error: TTS not initialized"
+
             text_subject: Subject[str] = Subject()
             audio_complete = threading.Event()
             self._tts_node.consume_text(text_subject)

dimos/agents/skills/speak_skill_spec.py

@@ -0,0 +1,21 @@
+# Copyright 2026 Dimensional Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Protocol
+
+from dimos.spec.utils import Spec
+
+
+class SpeakSkillSpec(Spec, Protocol):
+    def speak(self, text: str, blocking: bool = True) -> str: ...

dimos/e2e_tests/conftest.py

@@ -125,3 +125,30 @@ def direct_cmd_vel_explorer() -> Generator[PersonTrackPublisher, None, None]:
     explorer.start()
     yield explorer
     explorer.stop()
+
+
+@pytest.fixture
+def explore_office(
+    direct_cmd_vel_explorer: DirectCmdVelExplorer,
+) -> Callable[[], None]:
+    points = [
+        (0, -7.07),
+        (-4.16, -7.07),
+        (-4.45, 1.10),
+        (-6.72, 2.87),
+        (-1.78, 3.01),
+        (-1.54, 5.74),
+        (3.88, 6.16),
+        (2.16, 9.36),
+        (4.70, 3.87),
+        (4.67, -7.15),
+        (4.57, -4.19),
+        (-0.84, -2.78),
+        (-4.71, 1.17),
+        (4.30, 0.87),
+    ]
+
+    def explore() -> None:
+        direct_cmd_vel_explorer.follow_points(points)
+
+    return explore

dimos/e2e_tests/test_patrol_and_follow.py

@@ -20,24 +20,6 @@
 from dimos.e2e_tests.conf_types import StartPersonTrack
 from dimos.e2e_tests.dimos_cli_call import DimosCliCall
 from dimos.e2e_tests.lcm_spy import LcmSpy
-from dimos.simulation.mujoco.direct_cmd_vel_explorer import DirectCmdVelExplorer
-
-points = [
-    (0, -7.07),
-    (-4.16, -7.07),
-    (-4.45, 1.10),
-    (-6.72, 2.87),
-    (-1.78, 3.01),
-    (-1.54, 5.74),
-    (3.88, 6.16),
-    (2.16, 9.36),
-    (4.70, 3.87),
-    (4.67, -7.15),
-    (4.57, -4.19),
-    (-0.84, -2.78),
-    (-4.71, 1.17),
-    (4.30, 0.87),
-]
 
 
 @pytest.mark.skipif_in_ci
@@ -48,7 +30,7 @@ def test_patrol_and_follow(
     start_blueprint: Callable[[str], DimosCliCall],
     human_input: Callable[[str], None],
     start_person_track: StartPersonTrack,
-    direct_cmd_vel_explorer: DirectCmdVelExplorer,
+    explore_office: Callable[[], None],
 ) -> None:
     start_blueprint(
         "--mujoco-start-pos",
@@ -66,12 +48,7 @@ def test_patrol_and_follow(
 
     time.sleep(5)
 
-    print("Starting discovery.")
-
-    # Explore the entire room by driving directly via /cmd_vel.
-    direct_cmd_vel_explorer.follow_points(points)
-
-    print("Ended discovery.")
+    explore_office()
 
     start_person_track(
         [

dimos/e2e_tests/test_security_module.py

@@ -0,0 +1,77 @@
+# Copyright 2026 Dimensional Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from collections.abc import Callable
+import time
+
+from dimos_lcm.std_msgs import String
+import pytest
+
+from dimos.e2e_tests.conf_types import StartPersonTrack
+from dimos.e2e_tests.dimos_cli_call import DimosCliCall
+from dimos.e2e_tests.lcm_spy import LcmSpy
+
+
+@pytest.mark.skipif_in_ci
+@pytest.mark.skipif_no_openai
+@pytest.mark.mujoco
+def test_security_module(
+    lcm_spy: LcmSpy,
+    start_blueprint: Callable[[str], DimosCliCall],
+    human_input: Callable[[str], None],
+    start_person_track: StartPersonTrack,
+    explore_office: Callable[[], None],
+) -> None:
+    start_blueprint(
+        "--mujoco-start-pos",
+        "-10.75 -6.78",
+        "--mujoco-camera-position",
+        "-0.797 0.007  0.468 26.825 88.998 -70.321",
+        "--nerf-speed",
+        "0.8",
+        "--dtop",
+        "run",
+        "--disable",
+        "spatial-memory",
+        "unitree-go2-security",
+    )
+
+    lcm_spy.save_topic("/rpc/McpClient/on_system_modules/res")
+    lcm_spy.save_topic("/security_state#std_msgs.String")
+    lcm_spy.wait_for_saved_topic("/rpc/McpClient/on_system_modules/res", timeout=120.0)
+
+    time.sleep(2)
+
+    explore_office()
+
+    start_person_track(
+        [
+            (-10.75, -6.78),
+            (0, -7.07),
+        ]
+    )
+    human_input(
+        "start the security patrol. Just call start_security_patrol. Do not ask me anything."
+    )
+
+    def predicate(s: String) -> bool:
+        return s.data == "FOLLOWING"
+
+    lcm_spy.wait_for_message_result(
+        "/security_state#std_msgs.String",
+        String,
+        predicate,
+        "Failed to transition to FOLLOWING.",
+        360,
+    )

dimos/experimental/security_demo/conftest.py

@@ -0,0 +1,87 @@
+# Copyright 2026 Dimensional Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import annotations
+
+import pytest
+
+from dimos.experimental.security_demo.security_module import SecurityModule
+from dimos.msgs.sensor_msgs.CameraInfo import CameraInfo
+from dimos.msgs.sensor_msgs.Image import Image
+from dimos.perception.detection.detectors.yolo import Yolo2DDetector
+from dimos.perception.detection.type.detection2d.bbox import Detection2DBBox
+from dimos.utils.data import get_data
+
+
+@pytest.fixture(scope="session")
+def yolo_detector():
+    detector = Yolo2DDetector(device="cpu")
+    yield detector
+    detector.stop()
+
+
+@pytest.fixture(scope="session")
+def person_image():
+    return Image.from_file(get_data("security_detection.png"))
+
+
+@pytest.fixture(scope="session")
+def empty_image():
+    return Image.from_file(get_data("security_no_detection.png"))
+
+
+@pytest.fixture()
+def security_module(mocker):
+    mocker.patch("dimos.experimental.security_demo.security_module._create_router")
+    mocker.patch("dimos.experimental.security_demo.security_module._create_visual_servo")
+    mocker.patch("dimos.experimental.security_demo.security_module.YoloPersonDetector")
+    mocker.patch("dimos.experimental.security_demo.security_module.EdgeTAMProcessor")
+
+    module = SecurityModule(camera_info=CameraInfo())
+
+    # Replace output streams with mocks for test assertions
+    module.detection = mocker.MagicMock()
+    module.security_state = mocker.MagicMock()
+    module.goal_request = mocker.MagicMock()
+    module.cmd_vel = mocker.MagicMock()
+
+    # These are set by framework wiring, not __init__
+    module._planner_spec = mocker.MagicMock()
+    module._speak_skill = mocker.MagicMock()
+
+    yield module
+
+    module.stop()
+
+
+@pytest.fixture()
+def make_detection(person_image):
+    def _make(
+        bbox=(100.0, 50.0, 300.0, 400.0),
+        track_id=1,
+        class_id=0,
+        confidence=0.9,
+        name="person",
+    ):
+        return Detection2DBBox(
+            bbox=bbox,
+            track_id=track_id,
+            class_id=class_id,
+            confidence=confidence,
+            name=name,
+            ts=0.0,
+            image=person_image,
+        )
+
+    return _make