feat(sim): add semantic runtime effects and parallelism

[yuecideng]

Description

Stack

• Layer: 5/14
• Base: feat/atomic-action-verified-runtime
• Depends on: feat(atomic-actions): add verified effect-aware runtime #495
• Followed by: feat(gym): add declarative expert program runtime #497
  Add the canonical SkillRuntime, physical evidence adapters, composite effect monitors, and versioned parallel schema/runtime. Semantic calls compile and execute through the existing atomic-action engine with row-local state, effects, retry, and traces.

Parallel execution remains fail closed: resource analysis and shared-clock coordination are implemented, but production concurrency still requires an authoritative physical safety validator.

Refs #471
Refs #474

Type of change

• New feature (non-breaking change which adds functionality)

Screenshots

Not applicable.

Validation

• Focused coverage: tests/sim/skills, including effects, evidence, runtime, parallel compiler/runtime, and conditional cuRobo recovery coverage
• Final affected-suite regression on the stack tip: 1215 passed, 2 skipped, 8 deselected
• Changed Python files pass Black 26.3.1; the Sphinx build and rollout-report drift check pass at the stack tip

Checklist

• Changed Python files pass Black 26.3.1.
• Corresponding public/design documentation is included in this stack.
• Tests cover the affected behavior.
• No dependency update is required.

[Copilot]

Pull request overview

Adds the canonical semantic-skill runtime layer (effects + evidence collection + runtime traces) and introduces a strict, versioned parallel-execution contract that compiles down to the existing atomic-action engine. This extends the simulation “skills” facade with first-class articulation semantics, effect-monitor configuration via policy presets, and deterministic metadata/tracing for downstream program/runtime layers.

Changes:

• Extend the scene/registry model to publish canonical articulation joint observations and evidence addresses.
• Add effect-monitor + evidence-source plumbing into robot skill profiles/presets and semantic integration validation.
• Introduce deterministic parallel timing/state-merge contracts plus extensive test coverage (including a GPU cuRobo dynamic-collision recovery gate).

Reviewed changes

Copilot reviewed 25 out of 25 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tests/sim/skills/test_scene.py	Adds joint-observation registration/snapshot behavior tests and simulation alias semantics coverage.
tests/sim/skills/test_runtime.py	New end-to-end tests for SkillRuntime behavior, tracing/metadata, row masking, and parallel factory interactions.
tests/sim/skills/test_profiles.py	Validates endpoint effect-source snapshotting and preset effect-monitor behavior.
tests/sim/skills/test_parallel.py	New tests for parallel alignment, claim conflicts, timing validation, and deterministic barrier/state merging.
tests/sim/skills/test_integration.py	Adds binding-time validation for safe preset dynamic-collision requirements and preset reachability semantics.
tests/sim/skills/test_evidence.py	New tests for evidence providers/collector registry, query building, and articulation evidence provider behavior.
tests/sim/skills/test_effects.py	New tests for effect contracts, monitor refs/registries, and composite monitor behavior.
tests/sim/skills/test_curobo_semantic_runtime_dynamic_recovery_gpu.py	GPU/cuRobo integration test ensuring replanning on dynamic collision-world updates.
tests/sim/skills/test_compiler.py	Extends compiler tests for effect monitor selection/validation and dynamic collision policy propagation.
tests/sim/skills/test_calls.py	Adds deterministic, JSON-safe semantic-call metadata serialization test.
tests/sim/skills/test_articulation_semantics.py	New tests for OperateArticulation call semantics, grounding, and typed articulation effects.
embodichain/lab/sim/skills/scene.py	Adds articulation joint evidence address + joint-state providers and integrates joint observations into scene versioning.
embodichain/lab/sim/skills/profiles.py	Adds endpoint effect-sources, task-state key plumbing, and preset-owned effect-monitor selections.
embodichain/lab/sim/skills/parallel.py	New deterministic parallel execution contracts (claims, timing grid, effect merge, barrier resolution).
embodichain/lab/sim/skills/integration.py	Adds OperateArticulation call linking and safe dynamic-collision policy preflight validation.
embodichain/lab/sim/skills/calls.py	Adds call metadata serialization and introduces OperateArticulation semantic call type.
embodichain/lab/sim/skills/init.py	Exposes new runtime/effects/evidence/parallel APIs through package exports.
docs/source/overview/sim/atomic_actions/robot_skill_profiles.md	Documents safe preset dynamic-collision rules and preset-driven effect monitor selection.
docs/source/api_reference/embodichain/embodichain.lab.sim.skills.rst	Expands API reference to include semantic runtime, effects/evidence, and parallel execution types.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[greptile-apps]

Greptile Summary

The PR adds a canonical semantic-skill runtime with typed physical-effect verification and a fail-closed parallel execution foundation.

• Replaces the earlier semantic task facade with per-call grounding, execution, evidence collection, tracing, and verified state transitions.
• Adds typed effect specifications, evidence providers, composite monitors, and row-local hysteresis.
• Adds resource and symbolic-write analysis, shared-clock branch coordination, command safety validation, and deterministic parallel state merging.
• Updates atomic-action primitives, public exports, documentation, and focused tests for the new contracts.

Confidence Score: 5/5

The PR appears safe to merge because no blocking failure remains in the eligible follow-up review scope.

No blocking failure remains.

Important Files Changed

Filename	Overview
embodichain/lab/sim/skills/runtime.py	Introduces the canonical sequential semantic runtime, per-call execution lifecycle, verified state propagation, structured failures, and compatibility facade.
embodichain/lab/sim/skills/effects.py	Adds typed semantic effects, evidence clauses, monitor selection, composite evaluation, and row-local hysteresis.
embodichain/lab/sim/skills/evidence.py	Adds versioned evidence routing, collection contexts, provider contracts, and backend-neutral physical observations.
embodichain/lab/sim/skills/parallel.py	Defines deterministic resource analysis, command alignment, barrier resolution, and symbolic state-merge rules.
embodichain/lab/sim/skills/parallel_runtime.py	Adds fail-closed shared-clock coordination for forked semantic lanes with mandatory outbound-command safety validation.
embodichain/lab/sim/skills/compiler.py	Extends semantic analysis and grounding to produce typed effect specifications, monitors, and parallel-analysis metadata.
embodichain/lab/sim/atomic_actions/state.py	Extends row-local symbolic task state and state-delta application for verified runtime effects.
tests/sim/skills/test_parallel_runtime.py	Adds focused coverage for synchronization, timeout, cancellation, safe-stop acknowledgement, command validation, and parallel state merging.

Sequence Diagram

sequenceDiagram
participant App
participant Runtime as SkillRuntime
participant Compiler as SemanticSkillCompiler
participant Runner as ExecutionRunner
participant Evidence as EffectEvidenceCollector
participant Monitor as EffectMonitor
App->>Runtime: start semantic calls
Runtime->>Compiler: analyze complete call window
loop Each executable call
Runtime->>Compiler: ground using fresh PlanningContext
Runtime->>Runner: execute atomic invocation
Runner-->>Runtime: request effect verification
Runtime->>Evidence: collect typed physical evidence
Evidence-->>Monitor: correlated row evidence
Monitor-->>Runtime: success/failure masks
Runtime->>Runtime: commit verified StateDelta
end
Runtime-->>App: SkillResult and traces

Loading

_{Reviews (6): Last reviewed commit: "wip" | Re-trigger Greptile}