Skip to content

Python API

Scope Description
Python API Functions used by the CLI; descriptions generated from docstrings
Installation Installation
Paths Configuration and .roboloom paths are relative to the project root

Public API

Module Contents Page
roboloom Session, EpisodeInfo, EpisodeResult, EpisodeFailed, SessionError, InProcessSession, load_config, resolve_config, init_project, Sample, the exceptions Session
roboloom.setup Measurement and setup steps (EEPROM, zeros, travel, gravity, poses, bench tests), Operator Setup
roboloom.policy The policy contract (Policy, PolicySpec), load_policy, hold, the LeRobot, remote and replay loaders Policy
roboloom.hooks HOOKS_API, Hook, Requirements, HookSpec, StepContext, ChunkPosition, events, HookChain Mock only Hooks
roboloom.recording Checking a dataset (check_dataset), episode records (episodes), the features (features) Recording

CLI and Python

Task CLI Python
Create a project roboloom init rakuda2 roboloom.init_project("rakuda2")
Align the EEPROM roboloom commission [robot] <node> setup.plan_commission(...), setup.apply_commission(plan, operator)
Measure the range of motion roboloom measure-travel <cal> --out ... setup.measure_travel(cal, out, operator)
Identify the gravity model roboloom identify-gravity <cal> --out ... --collect setup.identify_gravity(cal, out, operator, collect=True)
Check gravity compensation roboloom gravity-check [robot] leader right_arm --with gravity setup.gravity_check(...)
Tune gravity compensation roboloom tune-gravity [robot] leader right_arm --with gravity setup.open_tuning(...) → TuningSession
Record an episode roboloom record [robot] --task ... --seconds 30 Session.launch(...) → record(...)
Run with a policy roboloom infer [robot] --policy pkg:make --seconds 30 Session.launch(cfg, options=["policy"], policy="pkg:make") → record(...)
Record continuously from the operation screen roboloom run [robot] The Session.launch(...) status screen and start() / stop() / save()
Install the skill for coding agents roboloom skills install roboloom.agent_skills.install(root=".", targets=("agents", "claude"), force=False), agent_skills.check(...)
Reference Link
Command arguments CLI

Example

Workflow Link
Python session usage Session
Policy execution Running a policy
Evaluation Evaluation verdicts