Feather SDK Documentation

Documentation for SDK version 0.1.7.

Installation

The Feather SDK is a Python package (Python 3.9–3.14) for controlling Feather robots. Install it from PyPI:

pip install feathersdk

1. Set up your robot config if it doesn’t exist

Robot behavior is driven by a Python configuration file. The SDK ships starter configs for each supported robot revision — copy one into a working directory with the bundled helper. You only need to do this once:

python3 -m feathersdk.installation.copy_configs /feathersys/configs v0.5.1

Supported revisions are v0.4.0, v0.4.0_o-shaft, and v0.5.1. This creates /feathersys/configs/ containing robot_conf.py (the entry point) alongside the systems, services, and infra config modules it imports. Edit robot_conf.py to match your hardware.

Note

The destination directory must not already exist (the copy will not overwrite it) and must be writable — copying into a system path such as /feathersys/ may require elevated permissions. To copy elsewhere, pass any path you own, e.g. python3 -m feathersdk.installation.copy_configs ~/feather_configs v0.5.1, and use that path in the next step.

2. Initialize and control the robot

initialize_feather_system loads your config file and returns a ready-to-use FeatherRobot. Point it at the robot_conf.py you copied above:

from feathersdk.robot.robot import initialize_feather_system, FeatherRobot
from feathersdk.utils.feathertypes import cast

robot = cast(FeatherRobot, initialize_feather_system('/feathersys/configs/robot_conf.py'))

# Access the platforms and services enabled in your config
nav = robot.navigation_system
neck = robot.neck_system
torso = robot.torso_system
manipulation = robot.manipulation_system
data_monitoring = robot.data_monitor_service

Anything not present in your config is None, so only the systems and services you enabled are available.

More about configuration

Configs are executable Python. Running robot_conf.py builds a typed FeatherRobotConfig object and hands it to the SDK, so config values can be any Python expression and are validated when loaded.

File layout. copy_configs lays down a small tree. robot_conf.py is the entry point; it imports config objects built by the modules under infra/, services/, and systems/ and assembles them into one FeatherRobotConfig:

configs/
├── robot_conf.py            # entry point — assembles and exports the full config
├── infra/                   # comms_manager, data_monitor_service, logging
├── services/                # battery_service
└── systems/                 # navigation, neck, torso, arms, manipulation

How it’s loaded. The last line of robot_conf.py calls export_config(...). When you call initialize_feather_system('.../robot_conf.py') (or the lower-level load_configs), the SDK executes the file and reads back whatever you exported. Call export_config only once, in robot_conf.py; the other modules just build and return their config objects.

Enabling/disabling a system. Every system and service field is optional — set it to None to disable it (that robot.<name> attribute is then None at runtime). The shipped v0.5.1 config, for instance, sets manipulation_system_config=None and battery_service_config=None.

Editing safely. Leave version and set_root_feather_path(Path("/feathersys")) alone unless the hardware itself changed. Everything else — motor limits, tolerances, rates, safe-stop policy — is meant to be tuned. Each field is documented on its config class in the API Reference: FeatherRobotConfig, CommsManagerConfig, and the per-system *Config classes (e.g. ThreeWheelSwerveSystemConfig).

Package Overview

  • robot — robot control and platform systems (manipulation, navigation, neck, torso)

  • configs — Python-based configuration loading (load_configs, export_config)

  • comms — communication layer (CAN, network interfaces)

  • ipc — client/server IPC for driving a robot across processes

  • utils — shared helpers and type utilities

API Reference

API Reference

Release Notes

Release Notes