All pull requests

Pull requests

• docs(examples): hello-world prints the rollout summary

  #4865·maxmxf opened Oct 29, 2026·

  ·

• feat(v1): sim.light() adds a light, camera(name) re-aims a camera in place, describe() says how an arm reaches

  #4856·cagataycali opened Oct 28, 2026Member·

  ·
  5

• fix(v1): record= and a LeRobot checkpoint load without the encoder banner, progress bars or device warning

  #4855·cagataycali opened Oct 28, 2026Member·

  ·
  6

• fix(v1): a tool tells the model each joint's unit and range, size in metres, whole rpc sentences, the presets that drive it

  #4854·cagataycali opened Oct 28, 2026Member·

  ·
  5

• fix(v1): a LeRobot checkpoint's units are checked against its saved stats, not assumed

  #4848·cagataycali opened Oct 28, 2026Member·

  ·
  6

• feat(v1): Simulation("isaac") takes add_object(path=) files: a scene file is one body, free or static

  #4835·yinsong1986 opened Oct 27, 2026Contributor·

  ·

• fix(v1): a clock that raises never hides an e-stop, and Robot(clock=) refuses what is not a Clock instance

  #4809·rmncardoso opened Oct 19, 2026Collaborator·

  ·
  2

• fix(v1): command() refuses a NaN or wrong-length target before the body and clips into the limits, as act() does

  #4808·rmncardoso opened Oct 19, 2026Collaborator·

  ·
  1

• fix(v1): a replay keeps the norm mode each joint was recorded in, or a real body refuses it

  #4807·rmncardoso opened Oct 19, 2026Collaborator·

  ·
  1

• chore: name #4793's changelog fragment after its PR, so the release can assemble it

  #4805·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  1

• fix(v1): a real body refuses a LeRobot checkpoint that does not say its norm mode

  #4804·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  2

• fix(v1): a Feetech arm's calibration is the file Robot(id=) names, recorded as LeRobot records it; inf/nan hz refused; GLFW renders on macOS

  #4803·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  1

• fix(v1): PD gains survive every scene change, go on only after a run's checks, and are never NaN, infinite or negative

  #4802·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  13

• fix(v1): a state stream survives a failed read, reports it once and resumes at its rate

  #4801·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  2

• fix(v1): a twin never pushes a simulated arm or mirrors a NaN or huge reading; close() waits for the step in progress

  #4800·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  3

• fix(v1): a failed on_disconnect is retried, and the watchdog thread survives it

  #4799·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  1

• fix(v1): a tag on dev publishes 1.0.0rc1, and only the version the package reports

  #4796·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  1

• fix(v1): a corrupted feetech status packet is read again, not the end of a run or an e-stop's hold

  #4795·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  2

• fix(v1): while the e-stop is latched, a faulted run, a watchdog lapse and hold() write nothing to the body

  #4794·rmncardoso opened Oct 18, 2026Collaborator·

  ·
  2