Scheduler refactor [utils]: auto_importer, pydantic_utils, registry, singleton (#289)

## Summary

This PR introduces four new utility modules that provide foundational
capabilities for auto-discovery, registry patterns, singleton
management, and enhanced Pydantic utilities. It additionally
consolidates the pydantic utility classes from guidellm.objects into the
dedicated guidellm.utils module containing the new additions. This
change improves code organization and establishes the foundation for
upcoming scheduler refactoring work.

## Details

- **Moved core
utilities** from `guidellm.objects.pydantic` to `guidellm.utils.pydantic_utils`
- **Added `AutoImporterMixin`** for dynamic module discovery within
packages
- **Added `RegistryMixin`** for object registration with optional
auto-discovery
- **Added singleton
mixins** (`SingletonMixin`, `ThreadSafeSingletonMixin`) for instance
management
- **Enhanced Pydantic
utilities** with `ReloadableBaseModel`, `StandardBaseDict`,
and `PydanticClassRegistryMixin` for polymorphic model serialization
- **Updated all imports** across the codebase to reference new module
locations
- **Added comprehensive test coverage** for all new utility modules
- **Removed obsolete** `guidellm.objects.pydantic` module and associated
tests

## Test Plan

- Run existing test suite to ensure no regressions from import changes
- Execute new utility module test suites covering smoke, sanity, and
regression scenarios

## Related Issues


---

- [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 ##`)

---------

Signed-off-by: Mark Kurtz <[email protected]>
Co-authored-by: Copilot <[email protected]>

Author: markurtz

src/guidellm/backend/response.py

@@ -3,7 +3,7 @@
 from pydantic import computed_field
 
 from guidellm.config import settings
-from guidellm.objects.pydantic import StandardBaseModel
+from guidellm.utils import StandardBaseModel
 
 __all__ = [
     "RequestArgs",

src/guidellm/benchmark/aggregator.py

@@ -24,8 +24,6 @@
 from guidellm.config import settings
 from guidellm.objects import (
     RunningStats,
-    StandardBaseModel,
-    StatusBreakdown,
     TimeRunningStats,
 )
 from guidellm.request import (
@@ -40,7 +38,7 @@
     SchedulerRequestResult,
     WorkerDescription,
 )
-from guidellm.utils import check_load_processor
+from guidellm.utils import StandardBaseModel, StatusBreakdown, check_load_processor
 
 __all__ = [
     "AggregatorT",

src/guidellm/benchmark/benchmark.py

@@ -13,8 +13,6 @@
     ThroughputProfile,
 )
 from guidellm.objects import (
-    StandardBaseModel,
-    StatusBreakdown,
     StatusDistributionSummary,
 )
 from guidellm.request import (
@@ -32,6 +30,7 @@
     ThroughputStrategy,
     WorkerDescription,
 )
+from guidellm.utils import StandardBaseModel, StatusBreakdown
 
 __all__ = [
     "Benchmark",

src/guidellm/benchmark/benchmarker.py

@@ -22,7 +22,6 @@
 )
 from guidellm.benchmark.benchmark import BenchmarkArgs, GenerativeBenchmark
 from guidellm.benchmark.profile import Profile
-from guidellm.objects import StandardBaseModel
 from guidellm.request import (
     GenerationRequest,
     GenerativeRequestLoaderDescription,
@@ -37,6 +36,7 @@
     SchedulerRequestResult,
     SchedulingStrategy,
 )
+from guidellm.utils import StandardBaseModel
 
 __all__ = ["Benchmarker", "BenchmarkerResult", "GenerativeBenchmarker"]
 

src/guidellm/benchmark/output.py

@@ -23,13 +23,12 @@
 from guidellm.config import settings
 from guidellm.objects import (
     DistributionSummary,
-    StandardBaseModel,
     StatusDistributionSummary,
 )
 from guidellm.presentation import UIDataBuilder
 from guidellm.presentation.injector import create_report
 from guidellm.scheduler import strategy_display_str
-from guidellm.utils import Colors, split_text_list_by_length
+from guidellm.utils import Colors, StandardBaseModel, split_text_list_by_length
 
 __all__ = [
     "GenerativeBenchmarksConsole",

src/guidellm/benchmark/profile.py

@@ -5,7 +5,6 @@
 from pydantic import Field, computed_field
 
 from guidellm.config import settings
-from guidellm.objects import StandardBaseModel
 from guidellm.scheduler import (
     AsyncConstantStrategy,
     AsyncPoissonStrategy,
@@ -15,6 +14,7 @@
     SynchronousStrategy,
     ThroughputStrategy,
 )
+from guidellm.utils import StandardBaseModel
 
 __all__ = [
     "AsyncProfile",

src/guidellm/benchmark/scenario.py

@@ -11,8 +11,8 @@
 
 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.utils import StandardBaseModel
 
 __ALL__ = ["Scenario", "GenerativeTextScenario", "get_builtin_scenarios"]
 

src/guidellm/objects/__init__.py

@@ -1,4 +1,3 @@
-from .pydantic import StandardBaseModel, StatusBreakdown
 from .statistics import (
     DistributionSummary,
     Percentiles,
@@ -11,8 +10,6 @@
     "DistributionSummary",
     "Percentiles",
     "RunningStats",
-    "StandardBaseModel",
-    "StatusBreakdown",
     "StatusDistributionSummary",
     "TimeRunningStats",
 ]

src/guidellm/objects/pydantic.py

@@ -6,7 +6,7 @@
 import numpy as np
 from pydantic import Field, computed_field
 
-from guidellm.objects.pydantic import StandardBaseModel, StatusBreakdown
+from guidellm.utils import StandardBaseModel, StatusBreakdown
 
 __all__ = [
     "DistributionSummary",