[GuideLLM Refactor] scheduler package updates, rewrites, and tests expansion (#354)

## **Summary**

Introduces a comprehensive constraints system and enhanced timing
control for the scheduler refactor. The implementation moves from
hardcoded execution limits to a flexible, composable constraint system
that enables sophisticated benchmark stopping criteria. Additionally,
request timing calculations are moved from precalculated to per-request
basis, enabling dynamic rate adjustments and better distributed
coordination.

## **Details**

- **Added constraints system** (`constraints.py`): Implements
Protocol-based constraint architecture with support for request limits,
duration limits, error thresholds, and sliding window error rates
    - `MaxNumberConstraint`: Limits execution based on request count
    - `MaxDurationConstraint`: Limits execution based on time duration
- `MaxErrorsConstraint`: Limits execution based on absolute error count
- `MaxErrorRateConstraint`: Limits execution based on sliding window
error rate
- `MaxGlobalErrorRateConstraint`: Limits execution based on global error
rate
- `ConstraintsInitializerFactory`: Registry system for constraint
creation and serialization
- **Refactored core objects** (`objects.py`): Replaced `result.py` and
expanded capabilities
- Made scheduler package fully generic, decoupling from backend-specific
types
- Added `BackendInterface` protocol for type-safe backend integration
- Enhanced `ScheduledRequestInfo` with comprehensive timing and status
tracking
    - Added `SchedulerState` for distributed state coordination
- Introduced `SchedulerUpdateAction` for constraint-based control
signals
- **Enhanced scheduling strategies** (`strategy.py`): Introduced request
timing abstractions
- Added `ScheduledRequestTimings` base class for timing implementations
- `LastCompletionRequestTimings`: For synchronous and concurrent
strategies
    - `NoDelayRequestTimings`: For maximum throughput strategies
    - `ConstantRateRequestTimings`: For fixed-rate scheduling
    - `PoissonRateRequestTimings`: For stochastic request patterns
- Strategies now create per-worker timing instances instead of
precalculated schedules
- **Added environment abstractions** (`environment.py`): Coordination
layer for distributed execution
    - `Environment` protocol for distributed synchronization
- `NonDistributedEnvironment` implementation for single-node execution
- **Worker process management** (`worker.py`, `worker_group.py`):
Distributed request processing infrastructure
    - Individual worker process management with lifecycle coordination
    - Multi-process orchestration with state synchronization
    - Constraint evaluation and graceful shutdown coordination

## **Test Plan**

- Full unit tests and some integration tests added and passing

## **Related Issues**

- Part of scheduler refactor initiative to support distributed
benchmarking

---

- [x] "I certify that all code in this PR is my own, except as noted
below."

## **Use of AI**

- [x]  Includes AI-assisted code completion
- [x]  Includes code generated by an AI application
- [ ] Includes AI-generated tests (NOTE: AI written tests should have a
docstring that includes `## WRITTEN BY AI ##`)

Author: markurtz

src/guidellm/benchmark/scenario.py

@@ -12,7 +12,7 @@
 from guidellm.backend.backend import BackendType
 from guidellm.benchmark.profile import ProfileType
 from guidellm.objects.pydantic import StandardBaseModel
-from guidellm.scheduler.strategy import StrategyType
+from guidellm.scheduler.strategies import StrategyType
 
 __ALL__ = ["Scenario", "GenerativeTextScenario", "get_builtin_scenarios"]
 

src/guidellm/scheduler/__init__.py

@@ -1,47 +1,90 @@
-from .result import (
-    SchedulerRequestInfo,
-    SchedulerRequestResult,
-    SchedulerResult,
-    SchedulerRunInfo,
+from .constraints import (
+    Constraint,
+    ConstraintInitializer,
+    ConstraintsInitializerFactory,
+    MaxDurationConstraint,
+    MaxErrorRateConstraint,
+    MaxErrorsConstraint,
+    MaxGlobalErrorRateConstraint,
+    MaxNumberConstraint,
+    PydanticConstraintInitializer,
+    SerializableConstraintInitializer,
+    UnserializableConstraintInitializer,
+)
+from .environments import Environment, NonDistributedEnvironment
+from .objects import (
+    BackendInterface,
+    BackendT,
+    MeasuredRequestTimings,
+    MultiTurnRequestT,
+    RequestSchedulerTimings,
+    RequestT,
+    ResponseT,
+    ScheduledRequestInfo,
+    SchedulerMessagingPydanticRegistry,
+    SchedulerState,
+    SchedulerUpdateAction,
+    SchedulerUpdateActionProgress,
 )
 from .scheduler import Scheduler
-from .strategy import (
+from .strategies import (
     AsyncConstantStrategy,
     AsyncPoissonStrategy,
     ConcurrentStrategy,
+    ConstantRateRequestTimings,
+    LastCompletionRequestTimings,
+    NoDelayRequestTimings,
+    PoissonRateRequestTimings,
+    ScheduledRequestTimings,
     SchedulingStrategy,
+    StrategyT,
     StrategyType,
     SynchronousStrategy,
     ThroughputStrategy,
-    strategy_display_str,
-)
-from .worker import (
-    GenerativeRequestsWorker,
-    GenerativeRequestsWorkerDescription,
-    RequestsWorker,
-    ResolveStatus,
-    WorkerDescription,
-    WorkerProcessResult,
 )
+from .worker import WorkerProcess
+from .worker_group import WorkerProcessGroup
 
 __all__ = [
     "AsyncConstantStrategy",
     "AsyncPoissonStrategy",
+    "BackendInterface",
+    "BackendT",
     "ConcurrentStrategy",
-    "GenerativeRequestsWorker",
-    "GenerativeRequestsWorkerDescription",
-    "RequestsWorker",
-    "ResolveStatus",
+    "ConstantRateRequestTimings",
+    "Constraint",
+    "ConstraintInitializer",
+    "ConstraintsInitializerFactory",
+    "Environment",
+    "LastCompletionRequestTimings",
+    "MaxDurationConstraint",
+    "MaxErrorRateConstraint",
+    "MaxErrorsConstraint",
+    "MaxGlobalErrorRateConstraint",
+    "MaxNumberConstraint",
+    "MeasuredRequestTimings",
+    "MultiTurnRequestT",
+    "NoDelayRequestTimings",
+    "NonDistributedEnvironment",
+    "PoissonRateRequestTimings",
+    "PydanticConstraintInitializer",
+    "RequestSchedulerTimings",
+    "RequestT",
+    "ResponseT",
+    "ScheduledRequestInfo",
+    "ScheduledRequestTimings",
     "Scheduler",
-    "SchedulerRequestInfo",
-    "SchedulerRequestResult",
-    "SchedulerResult",
-    "SchedulerRunInfo",
+    "SchedulerMessagingPydanticRegistry",
+    "SchedulerState",
+    "SchedulerUpdateAction",
+    "SchedulerUpdateActionProgress",
     "SchedulingStrategy",
+    "SerializableConstraintInitializer",
+    "StrategyT",
     "StrategyType",
     "SynchronousStrategy",
     "ThroughputStrategy",
-    "WorkerDescription",
-    "WorkerProcessResult",
-    "strategy_display_str",
+    "UnserializableConstraintInitializer",
+    "WorkerProcess",
+    "WorkerProcessGroup",
 ]