Add fake demo runtime vertical slice

flashdreams/flashdreams/runtime/__init__.py

@@ -52,14 +52,20 @@
 from flashdreams.runtime.metrics import (
     InMemoryMetricsRecorder,
     MetricsRecorder,
+    MetricsSnapshot,
     NullMetricsRecorder,
     RuntimeMetricSample,
 )
 from flashdreams.runtime.output import NullOutputTarget, OutputArtifact, OutputTarget
 from flashdreams.runtime.runner import run_inference_session
-from flashdreams.runtime.types import StepRequest, StepResult
+from flashdreams.runtime.types import (
+    StepRequest,
+    StepRequirements,
+    StepResult,
+    step_requirements_from_request,
+)
 from flashdreams.runtime.video_output import Mp4VideoOutputTarget
-from flashdreams.runtime.worker import ThreadAffineRuntimeWorker
+from flashdreams.runtime.worker import ModelExecutionWorker, ThreadAffineRuntimeWorker
 
 __all__ = [
     "CanonicalInputs",
@@ -90,7 +96,9 @@
     "KeyboardToDriverCommand",
     "MappingCompatibility",
     "MetricsRecorder",
+    "MetricsSnapshot",
     "ModelAdapter",
+    "ModelExecutionWorker",
     "Mp4VideoOutputTarget",
     "NullMetricsRecorder",
     "NullOutputTarget",
@@ -100,10 +108,12 @@
     "RuntimeMetricSample",
     "ScriptedModality",
     "StepRequest",
+    "StepRequirements",
     "StepResult",
     "TimeWindow",
     "ThreadAffineRuntimeWorker",
     "run_inference_session",
+    "step_requirements_from_request",
     "undeclared_inference_inputs",
     "UserInputCapability",
     "UserInputEvent",

flashdreams/flashdreams/runtime/config.py

@@ -61,6 +61,9 @@ class InferenceConfig:
     cache_policy: str | None = None
     """Optional cache policy selector; ``None`` leaves the choice to the adapter."""
 
+    seed: int | None = None
+    """Optional seed used when resolving deterministic demo/runtime behavior."""
+
     runtime_options: Mapping[str, Any] = field(default_factory=dict)
     """Adapter/backend-specific runtime options."""
 
@@ -70,6 +73,11 @@ class InferenceConfig:
     def __post_init__(self) -> None:
         if not self.model_id.strip():
             raise ValueError("InferenceConfig.model_id must be non-empty.")
+        if self.seed is not None:
+            if isinstance(self.seed, bool) or not isinstance(self.seed, int):
+                raise TypeError("InferenceConfig.seed must be an integer.")
+            if self.seed < 0:
+                raise ValueError("InferenceConfig.seed must be >= 0.")
         object.__setattr__(
             self, "runtime_options", freeze_mapping(self.runtime_options)
         )

flashdreams/flashdreams/runtime/demo/__init__.py

@@ -3,28 +3,176 @@
 
 """Experimental shared demo API above the inference runtime API."""
 
-from flashdreams.runtime.demo.outputs import build_output_target
-from flashdreams.runtime.demo.replay import run_replay_demo
+from flashdreams.runtime.demo.drivers import (
+    CLEANUP_TIMEOUT_S,
+    BatchSessionDriver,
+    DriverInvariantError,
+    RealtimeSessionDriver,
+    run_demo_session,
+    run_demo_session_async,
+    shielded_session_cleanup,
+    uncancel_current_task,
+)
+from flashdreams.runtime.demo.host import (
+    ModelWarmupPlan,
+    RuntimeHost,
+    WarmupSessionInputs,
+)
+from flashdreams.runtime.demo.outputs import (
+    Mp4OutputSink,
+    NullOutputSink,
+    OutputDecision,
+    OutputSink,
+    SessionInfo,
+    build_output_sink,
+    build_output_target,
+)
+from flashdreams.runtime.demo.pipeline import StepOutcome, StepPipeline
+from flashdreams.runtime.demo.replay import OutputSinkFactory, run_replay_demo
+from flashdreams.runtime.demo.run_modes import (
+    AsyncSessionDriver,
+    BenchmarkErrorPolicy,
+    DefaultErrorPolicy,
+    ErrorAction,
+    InMemorySessionMetricsRecorder,
+    MetricsSnapshot,
+    Mp4ErrorPolicy,
+    NativeWindowErrorPolicy,
+    NoopTransportService,
+    NullErrorPolicy,
+    RunContext,
+    RunMode,
+    RunModeCapabilities,
+    RunModeWarmup,
+    RunResult,
+    RunSummary,
+    SessionDriver,
+    SessionEdges,
+    SingleSessionAdmissionPolicy,
+    WebRTCErrorPolicy,
+    build_model_warmup_plan,
+    warmup_run_context,
+)
+from flashdreams.runtime.demo.session_inputs import (
+    BatchInputSource,
+    ControlDecision,
+    InputSource,
+    ModelInputProvider,
+    PreparedStep,
+    ProviderCapabilities,
+    RealtimeInputSource,
+    UserInputWindow,
+)
 from flashdreams.runtime.demo.spec import (
     DemoAdapter,
     DemoSpec,
+    ModelWarmupAdapter,
     Mp4OutputSpec,
     NullOutputSpec,
     OutputSpec,
     PreparedScenario,
     WebRTCAppResources,
     WebRTCOutputSpec,
 )
+from flashdreams.runtime.demo.timing import (
+    SPARSE_KEY_SEGMENTS_METADATA_KEY,
+    ActivationPolicy,
+    ActivationResult,
+    ActivationSignal,
+    AlwaysActiveActivationPolicy,
+    CatchUpDecision,
+    CatchUpPolicy,
+    DeterministicClock,
+    KeyboardRealtimeInputSource,
+    RealtimeClock,
+    RealtimeWindowResult,
+    ResamplerRealtimeClock,
+    SignalActivationPolicy,
+    input_frame_count_from_request,
+)
+from flashdreams.runtime.demo.validation import (
+    ResolvedRunCapabilities,
+    resolve_run_capabilities,
+    validate_resolved_run,
+)
 
 __all__ = [
+    "BatchInputSource",
+    "BatchSessionDriver",
+    "CLEANUP_TIMEOUT_S",
+    "ControlDecision",
+    "DefaultErrorPolicy",
     "DemoAdapter",
     "DemoSpec",
+    "DriverInvariantError",
+    "ErrorAction",
+    "AsyncSessionDriver",
+    "ActivationPolicy",
+    "ActivationResult",
+    "ActivationSignal",
+    "AlwaysActiveActivationPolicy",
+    "BenchmarkErrorPolicy",
+    "InMemorySessionMetricsRecorder",
+    "InputSource",
+    "CatchUpDecision",
+    "CatchUpPolicy",
+    "DeterministicClock",
+    "MetricsSnapshot",
+    "ModelWarmupAdapter",
+    "ModelWarmupPlan",
+    "ModelInputProvider",
+    "KeyboardRealtimeInputSource",
+    "Mp4ErrorPolicy",
+    "Mp4OutputSink",
     "Mp4OutputSpec",
+    "NativeWindowErrorPolicy",
+    "NoopTransportService",
     "NullOutputSpec",
+    "NullOutputSink",
+    "NullErrorPolicy",
+    "OutputDecision",
+    "OutputSinkFactory",
     "OutputSpec",
+    "OutputSink",
     "PreparedScenario",
+    "PreparedStep",
+    "ProviderCapabilities",
+    "RealtimeInputSource",
+    "RealtimeClock",
+    "RealtimeSessionDriver",
+    "RealtimeWindowResult",
+    "ResolvedRunCapabilities",
+    "ResamplerRealtimeClock",
+    "RunContext",
+    "RunMode",
+    "RunModeCapabilities",
+    "RunModeWarmup",
+    "RunResult",
+    "RunSummary",
+    "RuntimeHost",
+    "SessionEdges",
+    "SessionDriver",
+    "SessionInfo",
+    "SignalActivationPolicy",
+    "SingleSessionAdmissionPolicy",
+    "SPARSE_KEY_SEGMENTS_METADATA_KEY",
+    "StepOutcome",
+    "StepPipeline",
+    "UserInputWindow",
+    "WarmupSessionInputs",
     "WebRTCAppResources",
+    "WebRTCErrorPolicy",
     "WebRTCOutputSpec",
+    "build_output_sink",
     "build_output_target",
+    "build_model_warmup_plan",
+    "input_frame_count_from_request",
+    "resolve_run_capabilities",
+    "run_demo_session",
+    "run_demo_session_async",
     "run_replay_demo",
+    "shielded_session_cleanup",
+    "uncancel_current_task",
+    "validate_resolved_run",
+    "warmup_run_context",
 ]

flashdreams/flashdreams/runtime/demo/app.py

@@ -6,6 +6,7 @@
 from __future__ import annotations
 
 import argparse
+import sys
 from abc import ABC, abstractmethod
 from typing import Any
 
@@ -29,10 +30,18 @@ def main(self, argv: list[str] | None = None) -> None:
         configure_logging()
         args = self.parse_args(argv)
         if args.command == "replay":
-            run_replay_demo(
+            result = run_replay_demo(
                 spec=self.replay_spec(args),
                 adapter=self.replay_adapter(),
             )
+            if result.status != "completed":
+                reason = result.reason or (
+                    str(result.error) if result.error is not None else None
+                )
+                if reason is None:
+                    reason = f"Replay demo ended with status {result.status!r}."
+                print(reason, file=sys.stderr)
+                raise SystemExit(1)
             return
         if args.command == "webrtc":
             context = initialize_cuda_distributed(