Skip to content

wandelbotsgmbh/wandelbots-nova / .agents/skills/nova-api-v2

nova-api-v2

Verified

Authoritative reference for working with the released Wandelbots NOVA API v2 in the wandelbots-nova Python SDK. Use whenever a task touches the NOVA API, the generated `wandelbots_api_client` models (`nova.api.models.*`), motion planning requests (`plan_trajectory`, `MotionGroupSetup`, `LimitSet`, `LimitsOverride`), robot limits, or torque/velocity/acceleration/jerk limits. Reach for this skill before guessing field names or assuming a field exists — the installed client reflects the RELEASED API, while nightly specs may contain unreleased fields. Especially relevant for controller-reported limits (e.g. KUKA torque) and any "does field X exist / how is it forwarded to the planner?" question.

NOVA API v2

Domain knowledge for the released Wandelbots NOVA API v2 as used by the wandelbots-nova SDK. The goal is to keep you from guessing field names or inventing fields that only exist in unreleased/nightly specs.

Source of truth (in priority order)

  1. The installed client wandelbots_api_client (pinned in pyproject.toml, e.g. wandelbots_api_client~=26.4.0). nova.api lazily re-exports wandelbots_api_client.v2_pydantic; all models live in a single generated file wandelbots_api_client/v2_pydantic/models/models.py. This matches the released API exactly.
  2. Docs UI: https://portal.wandelbots.io/docs/api/v2/ui/ — base URL /api/v2. Shows the API version (e.g. v2.4.0).

Prefer the installed models over any nightly/internal spec. A nightly spec may advertise fields that are not yet released and therefore are not in the client the SDK ships with.

How to verify a field exists (do this instead of guessing)

# Confirm versions
./.venv/bin/python -c "import wandelbots_api_client as c; print(c.__version__)"

# Inspect a model's fields directly
./.venv/bin/python -c "from nova import api; print(api.models.JointLimits.model_fields.keys())"

You can also grep the generated models file:

grep -n "^class JointLimits" .venv/lib/python*/site-packages/wandelbots_api_client/v2_pydantic/models/models.py

If a field is not in the installed model, treat it as not available in the released API, regardless of what a spec or ticket claims.

Limits & torque (motion planning)

This is the most common source of confusion. The released v2 API distinguishes global motion-group limits from per-segment limit overrides, and they do NOT carry the same fields.

  • The controller reports limits in MotionGroupDescription.operation_limits (auto_limits / manual_limits / manual_t1_limits / manual_t2_limits), each a LimitSet.
  • Per-joint limits live in LimitSet.joints[i] (JointLimits): position, velocity, acceleration, jerk, torque.
  • torque is a global per-joint limit only. There is no per-segment torque field.
  • MotionGroupSetup.global_limits (sent to plan_trajectory) is built from auto_limits in nova/utils/collision_setup.py::motion_group_setup_from_motion_group_description. So any controller-reported joints[i].torque (e.g. from a KUKA controller) is already forwarded to the planner unchanged via global_limits — no SDK change is needed for forwarding.
  • Per-segment MotionCommand.limits_override (LimitsOverride) has no torque field — only velocity / acceleration / jerk for joints and TCP. Therefore torque cannot be overridden per motion; it is exclusively a motion-group-global limit.

Consequence for MotionSettings

nova.types.motion_settings.MotionSettings is per-motion. Because the API only supports torque as a global limit, exposing a per-motion torque setting would be a footgun: it could only take effect on the collision-free planning path (which folds settings into global_limits via update_motion_group_setup_with_motion_settings) and would be silently ignored on the standard collision-checked path (which uses LimitsOverride). Do not add per-motion torque unless the released API gains a per-segment torque field.

Planning data flow (where to look)

  • Setup is built in nova/utils/collision_setup.py::motion_group_setup_from_motion_group_description.
  • Collision-checked planning: MotionGroup._plan_with_collision_check → plan_trajectory (uses LimitsOverride per MotionCommand, plus global_limits).
  • Collision-free planning: MotionGroup._plan_collision_free → plan_collision_free (folds MotionSettings into global_limits, not via LimitsOverride).
  • Per-segment overrides are capped to the global maxima before planning by nova/utils/motion_group_settings.py::clamp_motion_commands_to_global_limits (this prevents the planner from producing a trajectory the hardware down-scales at execution time).

Verifying behavior with a quick experiment

When unsure whether a value survives setup-building and serialization, construct the models and serialize the request rather than reasoning about it:

from nova import api
from nova.utils.collision_setup import motion_group_setup_from_motion_group_description as build

desc = api.models.MotionGroupDescription(
    motion_group_model=api.models.MotionGroupModel("UniversalRobots_UR5e"),
    operation_limits=api.models.OperationLimits(
        auto_limits=api.models.LimitSet(
            joints=[api.models.JointLimits(velocity=3.0, acceleration=10.0, torque=150.0)],
            tcp=api.models.CartesianLimits(velocity=250.0),
        )
    ),
)
setup = build(desc)
print(setup.global_limits.joints[0].torque)  # -> 150.0, i.e. forwarded

Pitfalls

  • Don't trust a ticket/spec that says a field exists until you've confirmed it in the installed model. Released ≠ nightly.
  • global_limits (request input) vs operation_limits (controller-reported description) are different objects — the setup builder copies auto_limits into global_limits.
  • Physical limit enforcement happens at execution time (the motion player time-scales the path); the planner treats limits_override as a per-segment replacement, not a clamp.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 465kUpdated Claude CodeCodexVerifiedInstall
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 465kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 318kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 318kUpdated Claude CodeCodexVerifiedInstall