Align motion API with AGENTS.md recommendations

- Use goto_target() directly for the WAKING look (recommended primary API),
with body_yaw=None to leave body position unchanged after wake_up()
- Replace set_target(head=_look_direction(...)) with look_at_world(duration=0)
in the ACTIVE scan loop; duration=0 calls set_target() internally
- Remove hand-rolled _look_direction() helper that duplicated SDK rotation logic
- Drop scipy dependency (only used by _look_direction)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

pyproject.toml

@@ -13,7 +13,6 @@ dependencies = [
     "reachy-mini",
     "onnxruntime",
     "opencv-python-headless",
-    "scipy",
 ]
 keywords = ["reachy-mini-app", "reachy-mini"]
 

recognizer/main.py

@@ -107,7 +107,8 @@ class Recognizer(ReachyMiniApp):
 
                 # Look toward the sound: DoA 0=left π/2=front π=right → world-y
                 y = math.sin(doa_angle - math.pi / 2) * 0.6
-                reachy_mini.look_at_world(1.0, y, 0.0, duration=0.5)
+                head_pose = reachy_mini.look_at_world(1.0, y, 0.0, perform_movement=False)
+                reachy_mini.goto_target(head=head_pose, duration=0.5, body_yaw=None)
 
                 active_start = time.time()
                 scan_t0 = active_start
@@ -125,9 +126,7 @@ class Recognizer(ReachyMiniApp):
                 # Gentle head scan so the robot looks alive while waiting
                 elapsed = now - scan_t0
                 y_scan = math.sin(2 * math.pi * 0.15 * elapsed) * SCAN_AMPLITUDE
-                reachy_mini.set_target(
-                    head=_look_direction(1.0, y_scan, 0.0)
-                )
+                reachy_mini.look_at_world(1.0, y_scan, 0.0, duration=0)
 
                 # Throttled face recognition
                 if now - last_face_check >= FACE_INTERVAL:
@@ -183,27 +182,6 @@ class Recognizer(ReachyMiniApp):
         reachy_mini.goto_sleep()
 
 
-def _look_direction(x: float, y: float, z: float) -> np.ndarray:
-    """Return a 4×4 head-pose matrix that points the head toward (x, y, z)."""
-    from scipy.spatial.transform import Rotation as R
-
-    target = np.array([x, y, z], dtype=float)
-    target /= np.linalg.norm(target)
-    forward = np.array([1.0, 0.0, 0.0])
-
-    axis = np.cross(forward, target)
-    axis_norm = np.linalg.norm(axis)
-    if axis_norm < 1e-8:
-        rot_mat = np.eye(3) if np.dot(forward, target) > 0 else R.from_euler("y", 180, degrees=True).as_matrix()
-    else:
-        angle = np.arccos(np.clip(np.dot(forward, target), -1.0, 1.0))
-        rot_mat = R.from_rotvec(angle * axis / axis_norm).as_matrix()
-
-    pose = np.eye(4)
-    pose[:3, :3] = rot_mat
-    return pose
-
-
 if __name__ == "__main__":
     app = Recognizer()
     try: