Secondary Operations

Secondary operations live in src/secondary_operations/ as single-file, lightweight processing steps. They must extend OperationInstance (same as main operations) and implement a run() method.

When to use secondary operations

• Operation logic fits in a single file (under ~200 lines).
• No heavy model loading, no separate module directory needed.
• Source is added and edited in the repository; the WebUI does not edit operation source.

Pattern

from typing import Any
from src.main_operations.definitions.base.base_class import OperationInstance

class MySecondaryOp(OperationInstance):
    def __init__(self, threshold: float = 0.5) -> None:
        self.threshold = threshold

    def run(self, input_data: Any) -> Any:
        # Transform input_data and return output
        return input_data

The class name is CamelCase derived from the snake_case file name, such as tag_filter.py to TagFilter.

Bundled secondary operations

File	Class	Purpose
angle_to_objects.py	AngleToObjects	Calculates horizontal angles to detections
camera_adjust.py	CameraAdjust	Applies camera brightness, contrast, saturation, gain, and exposure settings
camera_local_to_robot_transform.py	CameraLocalToRobotTransform	Converts camera-local detections to robot coordinates
camera_pose_output.py	CameraPoseOutput	Sends camera poses to the WebUI
camera_to_robot_pose.py	CameraToRobotPose	Applies camera extrinsics to produce robot pose
detected_objects_output.py	DetectedObjectsOutput	Sends detected objects to the WebUI
device_input.py	DeviceInput	Reads frames from a camera by camera_bus_id
extract_pose.py	ExtractPose	Extracts 2D pose data from a transform
flatten_pose.py	FlattenPose	Removes height and 3D rotation components
get_networktables_value.py	GetNetworktablesValue	Reads a NetworkTables value
ground_plane_intersection.py	GroundPlaneIntersection	Projects detections onto the ground plane
minimum_apriltag_count.py	MinimumApriltagCount	Rejects frames with too few AprilTags
pose_fusion.py	PoseFusion	Combines pose estimates with outlier rejection
pose_outlier_filter_rust.py	PoseOutlierFilterRust	Filters pose outliers using predictive gating
publish_to_networktables.py	PublishToNetworktables	Publishes typed values to NetworkTables
robot_local_to_field_transform.py	RobotLocalToFieldTransform	Converts robot-local detections to field coordinates
robot_pose_output.py	RobotPoseOutput	Sends robot pose to the WebUI
tag_filter.py	TagFilter	Includes or excludes detections by AprilTag ID

Config definitions

Each secondary operation needs a matching config definition:

src/secondary_operations/config_data/<name>_config_def.json

For example, minimum_apriltag_count_config_def.json names the class, ports, category, and settings used by the pipeline editor:

{
  "class_name": "MinimumApriltagCount",
  "category": "filt",
  "input_nodes": [{"name": "detections", "has_default": false}],
  "output_nodes": ["detections"],
  "parameters": {
    "minimum_detections": {
      "type": "int",
      "default": 2,
      "required": false
    }
  }
}

Promoting to a main operation

If a secondary operation grows complex enough to warrant its own module or requires separate test coverage, move it to src/main_operations/definitions/<name>.py (rename the class to <Name>Definition) and create an implementation module under src/main_operations/modules/<name>/.