refactor(atomic-actions): add typed tracking contracts

[yuecideng]

Description

Stack

• Layer: 9/14
• Base: feat/expert-program-rollout-validation
• Depends on: feat(tasks): add declarative task slices and rollout validation #500
• Followed by: feat(expert-program): add pre-simulation runtime catalog #504
  Replace implicit trajectory-error callbacks with typed tracking observations, metric bands, in-flight/terminal policies, and immutable tracking results.

The execution session can now diagnose row-local divergence, success, staleness, and retry through a stable public contract shared by direct atomic actions and semantic runtimes.

Refs #471
Refs #474

Type of change

• Breaking change (tracking callbacks are replaced by typed contracts)

Screenshots

Not applicable.

Validation

• Focused coverage: tests/sim/atomic_actions/test_tracking.py, test_engine_per_env.py, test_runner.py, and semantic runtime/profile tests
• 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.

[greptile-apps]

Greptile Summary

The PR replaces implicit trajectory-error feedback with immutable typed tracking contracts shared by direct atomic actions and semantic skill runtimes.

• Adds typed feedback sources, channel bindings, projectors, metrics, evaluators, policies, observations, and tracking results.
• Projects command frames into endpoint-local tracking setpoints and validates tracking contracts during planning.
• Updates execution to handle in-flight divergence, invalid feedback, terminal settling, consecutive acceptance, and bounded retries.
• Propagates tracking policies and bindings through atomic-action and semantic-skill compilation/runtime paths.
• Adds focused coverage for tracking contracts, per-environment execution, profiles, compilation, and runtime behavior.

Confidence Score: 5/5

The PR appears safe to merge; no concrete blocking failure or independently actionable non-blocking issue remains.

The typed tracking contracts are propagated through planning, bindings, semantic compilation, and execution with explicit validation, fail-closed feedback handling, and lifecycle resets across retries and replans.

Important Files Changed

Filename	Overview
embodichain/lab/sim/atomic_actions/tracking.py	Introduces validated immutable tracking addresses, observations, channel bindings, projectors, metrics, evaluators, policies, registries, and runtime aggregation.
embodichain/lab/sim/atomic_actions/execution.py	Replaces joint-specific error handling with typed in-flight and terminal evaluation, fail-closed feedback handling, settling, and retry lifecycle management.
embodichain/lab/sim/atomic_actions/core.py	Projects transport-neutral command frames through endpoint-owned tracking channels and rejects plans missing required metric channels.
embodichain/lab/sim/atomic_actions/plans.py	Replaces the feedback-mode enum with validated tracking policies and owned timed tracking sequences.
embodichain/lab/sim/atomic_actions/bindings.py	Adds immutable endpoint-local tracking-channel bindings with endpoint-address consistency checks.
embodichain/lab/sim/skills/profiles.py	Adds tracking policies to semantic presets and typed joint-position tracking bindings to control-part endpoints.
embodichain/lab/sim/skills/compiler.py	Preserves semantic profile tracking policies when grounding calls into atomic-action invocations.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    A[Action invocation] --> B[Resolved tracking policy]
    B --> C[Plan command frames]
    C --> D[Project endpoint tracking setpoints]
    D --> E[Execute command frame]
    E --> F[Read typed feedback]
    F --> G[Evaluate configured metrics]
    G -->|In flight accepted| E
    G -->|Diverged| H[Replan or exhaust recovery]
    G -->|Invalid feedback| I[Fail affected rows closed]
    E --> J[Terminal acceptance]
    J -->|Pending| K[Maintain terminal command]
    K --> F
    J -->|Accepted| L[Complete action]
    J -->|Timeout| M[Retry or exhaust action]

Loading

_{Reviews (1): Last reviewed commit: "refactor(atomic-actions): add typed trac..." | Re-trigger Greptile}

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.