Unify serving paths

[gtong-nv]

No description provided.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[greptile-apps]

Greptile Summary

The PR consolidates replay, WebRTC, and local-window serving around common runtime, result, output, and lifecycle abstractions.

• Introduces a canonical layout-aware StepResult and separates video collection from post-processing.
• Adds a thread-affine runtime worker and migrates integration lifecycle operations onto it.
• Unifies WebRTC management, encoding, application construction, and capability-driven output routing.
• Reuses shared Lingbot and OmniDreams model-session cores across their supported output modes.

Confidence Score: 5/5

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

No blocking failure remains.

Important Files Changed

Filename	Overview
flashdreams/flashdreams/infra/results.py	Defines the canonical immutable, layout-aware inference result consumed across output paths.
flashdreams/flashdreams/infra/video_output.py	Refactors post-processing to emit StepResult objects and moves collection into a dedicated collector.
flashdreams/flashdreams/runtime/worker.py	Adds serialized thread-affine runtime execution with ordered draining during shutdown.
flashdreams/flashdreams/serving/webrtc/manager.py	Migrates WebRTC session orchestration to the canonical request/result runtime boundary.
flashdreams/flashdreams/serving/output_targets.py	Replaces model-name routing with integration-provided output adapter capabilities.
integrations/lingbot/lingbot/model_session.py	Centralizes Lingbot generation and output-stream state for reuse across serving modes.
integrations/omnidreams/omnidreams/model_session.py	Centralizes OmniDreams model lifecycle and output processing across replay, WebRTC, and local presentation.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  Inputs["Replay / WebRTC / Local inputs"] --> Runtime["Thread-affine runtime worker"]
  Runtime --> Session["Integration model session"]
  Session --> Stream["VideoOutputStream"]
  Stream --> Result["StepResult"]
  Result --> MP4["MP4 output target"]
  Result --> WebRTC["WebRTC encoder"]
  Result --> Local["Local presentation"]

Loading

_{Reviews (3): Last reviewed commit: "Share demo application lifecycle" | Re-trigger Greptile}