pynetft 2.0.0

Python client for ATI Net F/T Ethernet force/torque sensors

pyNetFT

pyNetFT is a synchronous, typed Python client for ATI Industrial Automation Net F/T Ethernet force/torque sensors. It discovers the sensor calibration, streams RDT measurements through a native C++ core, and exposes raw counts, physical measurements, health, and recovery information without requiring NumPy.

• Tested native core: the protocol and recovery implementation is pinned to netft-cpp 0.2.2, includes its GCC 10/manylinux2014 compatibility fix and corrected libcurl minimum, and is exercised with offline fake sensors and native sanitizers.
• Sensor-aware data: calibration, force and torque units, configuration revisions, sequence progress, and faults remain visible to the application.
• Self-contained typed wheels: Linux wheels include the native core and a minimal static HTTP-only curl build, plus inline type information for Python 3.10–3.14.

Supported platforms

Installation	Python	Platform	Architecture	Support
PyPI wheel	CPython 3.10–3.14	manylinux2014 or newer Linux	x86_64	Supported
PyPI wheel	CPython 3.10–3.14	manylinux2014 or newer Linux	AArch64	Supported
Source build	CPython 3.10–3.14	POSIX	Host compiler architecture	Best effort

Windows, macOS wheels, PyPy, and asyncio are not supported by the 2.x package.

Installation

Install a supported wheel from PyPI:

python -m pip install pynetft

For a source build, use a C++17 compiler, CMake 3.16 or newer, and libcurl 7.63.0 or newer:

git clone https://github.com/netft/pyNetFT.git
cd pyNetFT
python -m pip install .

Source builds use the checked-in native core and do not download netft-cpp.

Quick start

The address below is the ATI factory default. Replace it with the address assigned to your sensor.

from pynetft import Client, Config

config = Config(sensor_host="192.168.1.1")

with Client(config) as client:
    sample = next(client.samples(timeout=1.0))
    print(sample.force, sample.force_unit.value)
    print(sample.torque, sample.torque_unit.value)

Client.samples() provides blocking iterator delivery. A read timeout raises TimeoutError, while a terminal sensor or transport fault raises SensorFaultError with structured fault and health data. Applications that integrate with a callback-driven loop can instead pass callback= to Client; the two delivery modes are mutually exclusive for a client run. See the API reference for lifecycle, callback, timeout, health, and error behavior.

Units and raw counts

Sample.force and Sample.torque preserve the units reported by the sensor, including the common ATI combination of newtons and newton-millimeters. pyNetFT does not apply ROS-specific SI conversion. Sample.wrench is the six-element physical-value tuple (Fx, Fy, Fz, Tx, Ty, Tz), while Sample.raw_wrench contains the six signed integer counts from the RDT record.

Automatic discovery is the default and should be preferred when the sensor configuration is authoritative. A complete Calibration override is available when independently verified fixed values are required.

Migrating from pyNetFT 1.x

The deprecated NetFT and Response interfaces remain available throughout 2.x. New code should use Client, Config, and Sample; see the 2.0 migration guide for exact call and field mappings.

Network security

ATI configuration discovery uses HTTP and RDT streaming uses unauthenticated UDP. Deploy the sensor and client on a trusted, isolated network segment, and do not expose either device protocol directly to the Internet or an untrusted shared network. See the security policy for the threat model and vulnerability reporting instructions.

Hardware safety

Force/torque measurements can affect motion and protective limits. Validate discovered calibration and units before enabling a controller, detect stale or faulted data, and retain an independent safety-rated control path. Client.bias() changes the sensor's software bias; call it only when the sensor is unloaded and the operation is explicitly authorized.

Contributing

Bug reports, hardware compatibility reports, documentation improvements, and focused code contributions are welcome. The contribution guide covers the Pixi environment, offline checks, hardware authorization, and native-core synchronization rules.

License

pyNetFT is licensed under the Apache License 2.0. The previous pyNetFT MIT license, the synchronized netft-cpp license, and the curl license notice are retained with the source and distribution artifacts.