dexi-rs 0.3.0

Fast Rust hand retargeting with Python bindings

dexi-rs

dexi-rs is a fast hand-retargeting package backed by a Rust engine. It loads YAML config files and bundled robot-hand URDF assets, then maps 3D human-hand targets to robot-hand joint positions.

Install distribution dexi-rs; import module dexi_rs.

uv pip install dexi-rs

Quickstart

import dexi_rs

cfg = dexi_rs.load_config("configs/teleop/allegro_hand_right.yml")
retargeting = cfg.build()

# Four wrist-to-fingertip vectors for this Allegro vector config.
target = [
    0.03, -0.02, 0.08,
    0.04, 0.00, 0.09,
    0.03, 0.02, 0.085,
    0.02, 0.04, 0.07,
]
qpos = retargeting.retarget(target)
print(dict(zip(retargeting.joint_names, qpos)))

Resource helpers

import dexi_rs

print(dexi_rs.available_configs("configs"))
config_path = dexi_rs.config_path("configs/teleop/allegro_hand_right.yml")
asset_root = dexi_rs.asset_path("robots/hands")

Retargeting API

• RetargetingConfig.from_file(path) loads a YAML config.
• dexi_rs.load_config(path) loads a config file from a filesystem path.
• config.build() creates a stateful SeqRetargeting instance.
• retargeting.retarget(ref_value, fixed_qpos=None) returns full robot qpos in retargeting.joint_names order.
• retargeting.reset(), set_qpos(), and get_qpos() control sequence state.
• retargeting.link_positions(qpos, link_names) returns world-frame 3D points for visualization.

Repository config families cover Allegro, Shadow, Schunk SVH, LEAP, Ability, Inspire, and Panda gripper hands for position, vector, and DexPilot-style retargeting where available.