feat(runtime): basic PolicyRuntime implementation

[maxxgx]

Summary

This PR contains Phase 1 (of 2):

• Initial policy runtime and async execution (threaded).
• Thread safe action queue
• Chunk smoothing: replace or merge via lerp (linear interpolation)

Phase 2: #119

Why

Validation

Breaking changes

• None

Related issues

• Closes #

[github-advanced-security]

You are seeing this message because GitHub Code Scanning has recently been set up for this repository, or this pull request contains the workflow file for the Code Scanning tool.

What Enabling Code Scanning Means:

• The 'Security' tab will display more code scanning analysis results (e.g., for the default branch).
• Depending on your configuration and choice of analysis tool, future pull requests will be annotated with code scanning analysis results.
• You will be able to see the analysis results for the pull request's branch on this overview once the scans have completed and the checks have passed.

For more information about GitHub Code Scanning, check out the documentation.

[Copilot]

Pull request overview

Introduces a new physicalai.runtime package implementing PolicyRuntime, a control loop that drives a robot using a trained policy. It adds an ActionQueue with pluggable ChunkSmoothers (Replace/Lerp), SyncExecution/AsyncExecution strategies (with worker health/watchdog and WorkerDiedError propagation), and a RuntimeCallback Protocol for hooks. Supporting changes fix InferenceModel.use_action_queue / select_action / predict_action_chunk so both single-pass and chunking runners work uniformly, expose RobotObservation, and add an async reference example.

Changes:

• New physicalai.runtime package: PolicyRuntime, ActionQueue, SyncExecution/AsyncExecution, ChunkSmoother/LerpSmoother/ReplaceSmoother.
• Reworked InferenceModel.predict_action_chunk / select_action to support both runner types, with shape normalization.
• Comprehensive unit tests for smoothers, action queue (incl. thread safety), execution strategies, and runtime callbacks.

Reviewed changes

Copilot reviewed 13 out of 14 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
src/physicalai/runtime/runtime.py	New PolicyRuntime, RunStats, default obs converter, callback dispatch.
src/physicalai/runtime/execution.py	Execution ABC, SyncExecution, threaded AsyncExecution with watchdog.
src/physicalai/runtime/_action_queue.py	Thread-safe queue with smoother-driven push_chunk.
src/physicalai/runtime/smoothers.py	Replace/Lerp chunk smoothers + input validation.
src/physicalai/runtime/init.py	Public runtime exports.
src/physicalai/inference/model.py	Unified select_action / predict_action_chunk; runner-based use_action_queue; chunk-shape normalization.
src/physicalai/robot/init.py	Re-exports RobotObservation.
tests/unit/runtime/*	New tests for smoothers, queue, execution, runtime.
tests/unit/inference/test_model.py	Updated tests reflecting non-raising chunk/select API.
examples/runtime/inference_async.py	Reference async inference example.
docs/design/components/runtime-system/implementation_plan.md	New implementation plan document.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[samet-akcay]

would be good to add your inference demo to our main readme file. Does not have to be in this pr, but we should have it :)