diff --git a/.github/release-please-config.json b/.github/release-please-config.json
index b6dd987b2a..ecc2145049 100644
--- a/.github/release-please-config.json
+++ b/.github/release-please-config.json
@@ -4,6 +4,7 @@
     ".": {
       "release-type": "python",
       "package-name": "ai-company",
+      "include-component-in-tag": false,
       "bump-minor-pre-major": true,
       "bump-patch-for-minor-pre-major": true,
       "changelog-sections": [
diff --git a/CLAUDE.md b/CLAUDE.md
index e99c953c7b..515ebab4bc 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -66,11 +66,11 @@ curl http://localhost:3000/api/v1/health   # backend (via web proxy)
 ```text
 src/ai_company/
   api/            # Litestar REST + WebSocket API (controllers, guards, channels)
-  budget/         # Cost tracking, budget enforcement (pre-flight/in-flight checks, auto-downgrade), billing periods, cost tiers, quota/subscription tracking, CFO cost optimization (anomaly detection, efficiency analysis, downgrade recommendations, approval decisions), spending reports
+  budget/         # Cost tracking, budget enforcement (pre-flight/in-flight checks, auto-downgrade), billing periods, cost tiers, quota/subscription tracking, CFO cost optimization (anomaly detection, efficiency analysis, downgrade recommendations, approval decisions), spending reports, budget errors (BudgetExhaustedError, DailyLimitExceededError, QuotaExhaustedError)
   cli/            # CLI interface (future — thin API wrapper if needed)
   communication/  # Message bus, dispatcher, messenger, channels, delegation, loop prevention, conflict resolution, meeting protocol
   config/         # YAML company config loading and validation
-  core/           # Shared domain models and base classes
+  core/           # Shared domain models, base classes, and resilience config (RetryConfig, RateLimiterConfig)
   engine/         # Agent orchestration, execution loops, parallel execution, task decomposition, routing, task assignment, task lifecycle, recovery, shutdown, workspace isolation, coordination error classification, and prompt policy validation
   hr/             # HR engine: hiring, firing, onboarding, offboarding, agent registry, performance tracking (task metrics, collaboration scoring, trend detection), promotion/demotion (criteria evaluation, approval strategies, model mapping)
   memory/         # Persistent agent memory (Mem0 initial, custom stack future — ADR-001), retrieval pipeline (ranking, injection, context formatting, non-inferable filtering), shared org memory (org/), consolidation/archival (consolidation/)
diff --git a/DESIGN_SPEC.md b/DESIGN_SPEC.md
index d362aebfc7..c489d8a459 100644
--- a/DESIGN_SPEC.md
+++ b/DESIGN_SPEC.md
@@ -2775,7 +2775,8 @@ ai-company/
 │       │   ├── artifact.py         # Produced work items
 │       │   ├── role.py             # Role model
 │       │   ├── role_catalog.py     # Role catalog
-│       │   └── personality.py     # Personality compatibility scoring
+│       │   ├── personality.py     # Personality compatibility scoring
+│       │   └── resilience_config.py # RetryConfig, RateLimiterConfig (shared by config.schema + providers.resilience)
 │       ├── engine/                  # Agent orchestration, execution loops, parallel execution, task decomposition, routing, task assignment, task lifecycle, recovery, shutdown, workspace isolation, coordination error classification, and prompt policy validation
 │       │   ├── errors.py           # Engine error hierarchy
 │       │   ├── prompt.py           # System prompt builder
@@ -3048,7 +3049,6 @@ ai-company/
 │       │   │   ├── router.py      # Router orchestrator
 │       │   │   └── strategies.py  # Routing strategies
 │       │   └── resilience/         # Resilience patterns
-│       │       ├── config.py      # RetryConfig, RateLimiterConfig
 │       │       ├── errors.py      # RetryExhaustedError
 │       │       ├── rate_limiter.py # Token bucket rate limiter
 │       │       └── retry.py       # RetryHandler with backoff
@@ -3157,6 +3157,7 @@ ai-company/
 │       │   ├── enums.py            # Budget-related enums
 │       │   ├── billing.py          # Billing period computation utilities
 │       │   ├── enforcer.py         # BudgetEnforcer service (pre-flight, in-flight, auto-downgrade)
+│       │   ├── errors.py           # BudgetExhaustedError, DailyLimitExceededError, QuotaExhaustedError
 │       │   ├── optimizer.py        # CostOptimizer service — anomaly detection, efficiency analysis, downgrade recommendations, approval decisions (M5)
 │       │   ├── optimizer_models.py # CostOptimizer domain models — anomaly, efficiency, downgrade, approval, config (M5)
 │       │   ├── quota.py            # Quota/subscription models, degradation config, quota snapshots
diff --git a/pyproject.toml b/pyproject.toml
index 73865b467e..aa14594df4 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -17,7 +17,7 @@ dependencies = [
     "aiosqlite==0.22.1",
     "jinja2==3.1.6",
     "jsonschema==4.26.0",
-    "litellm==1.82.0",
+    "litellm==1.82.1",
     "litestar[standard,structlog,pydantic,brotli,prometheus]==2.21.1",
     "mcp==1.26.0",
     "pydantic==2.12.5",
diff --git a/src/ai_company/budget/__init__.py b/src/ai_company/budget/__init__.py
index e73eb24303..2972039f27 100644
--- a/src/ai_company/budget/__init__.py
+++ b/src/ai_company/budget/__init__.py
@@ -38,6 +38,11 @@
 )
 from ai_company.budget.enforcer import BudgetEnforcer
 from ai_company.budget.enums import BudgetAlertLevel
+from ai_company.budget.errors import (
+    BudgetExhaustedError,
+    DailyLimitExceededError,
+    QuotaExhaustedError,
+)
 from ai_company.budget.hierarchy import (
     BudgetHierarchy,
     DepartmentBudget,
@@ -100,6 +105,7 @@
     "BudgetAlertLevel",
     "BudgetConfig",
     "BudgetEnforcer",
+    "BudgetExhaustedError",
     "BudgetHierarchy",
     "CategoryBreakdown",
     "CoordinationEfficiency",
@@ -113,6 +119,7 @@
     "CostTierDefinition",
     "CostTiersConfig",
     "CostTracker",
+    "DailyLimitExceededError",
     "DegradationAction",
     "DegradationConfig",
     "DepartmentBudget",
@@ -135,6 +142,7 @@
     "ProviderCostModel",
     "ProviderDistribution",
     "QuotaCheckResult",
+    "QuotaExhaustedError",
     "QuotaLimit",
     "QuotaSnapshot",
     "QuotaTracker",
diff --git a/src/ai_company/budget/enforcer.py b/src/ai_company/budget/enforcer.py
index fe7cabe93c..02c645b7ab 100644
--- a/src/ai_company/budget/enforcer.py
+++ b/src/ai_company/budget/enforcer.py
@@ -10,13 +10,13 @@
 
 from ai_company.budget.billing import billing_period_start, daily_period_start
 from ai_company.budget.enums import BudgetAlertLevel
-from ai_company.budget.quota import QuotaCheckResult
-from ai_company.constants import BUDGET_ROUNDING_PRECISION
-from ai_company.engine.errors import (
+from ai_company.budget.errors import (
     BudgetExhaustedError,
     DailyLimitExceededError,
     QuotaExhaustedError,
 )
+from ai_company.budget.quota import QuotaCheckResult
+from ai_company.constants import BUDGET_ROUNDING_PRECISION
 from ai_company.observability import get_logger
 from ai_company.observability.events.budget import (
     BUDGET_ALERT_THRESHOLD_CROSSED,
diff --git a/src/ai_company/budget/errors.py b/src/ai_company/budget/errors.py
new file mode 100644
index 0000000000..62d5220c1e
--- /dev/null
+++ b/src/ai_company/budget/errors.py
@@ -0,0 +1,32 @@
+"""Budget-layer error hierarchy.
+
+Defines budget-specific exceptions in a leaf module with no intra-project
+imports, preventing circular dependency chains when these exceptions are
+needed by both the budget enforcer and the engine layer.
+"""
+
+
+class BudgetExhaustedError(Exception):
+    """Budget exhaustion signal.
+
+    Used in two contexts:
+
+    1. Raised directly by :meth:`BudgetEnforcer.check_can_execute`
+       when pre-flight budget checks fail (monthly hard stop or daily
+       limit exceeded).
+    2. Caught by the engine layer (``AgentEngine.run``) and converted
+       into an ``ExecutionResult`` with
+       ``TerminationReason.BUDGET_EXHAUSTED``.
+    """
+
+
+class DailyLimitExceededError(BudgetExhaustedError):
+    """Per-agent daily spending limit exceeded."""
+
+
+class QuotaExhaustedError(BudgetExhaustedError):
+    """Raised when provider quota is exhausted.
+
+    Raised for all degradation strategies. Degradation routing
+    (FALLBACK/QUEUE) is tracked in M7.
+    """
diff --git a/src/ai_company/config/schema.py b/src/ai_company/config/schema.py
index 60baa7b5bb..39beb30bc2 100644
--- a/src/ai_company/config/schema.py
+++ b/src/ai_company/config/schema.py
@@ -18,6 +18,7 @@
     WorkflowHandoff,
 )
 from ai_company.core.enums import AutonomyLevel, CompanyType, SeniorityLevel
+from ai_company.core.resilience_config import RateLimiterConfig, RetryConfig
 from ai_company.core.role import CustomRole  # noqa: TC001
 from ai_company.core.types import NotBlankStr  # noqa: TC001
 from ai_company.hr.promotion.config import PromotionConfig
@@ -35,94 +36,6 @@
 logger = get_logger(__name__)
 
 
-# ── Resilience config models ─────────────────────────────────────
-# Defined here (not in providers.resilience) to avoid circular imports
-# between config ↔ providers. Re-exported by providers.resilience.config.
-
-
-class RetryConfig(BaseModel):
-    """Configuration for automatic retry of transient provider errors.
-
-    Attributes:
-        max_retries: Maximum number of retry attempts (0 disables retries).
-        base_delay: Initial delay in seconds before the first retry.
-        max_delay: Upper bound on computed delay in seconds.
-        exponential_base: Multiplier for exponential backoff.
-        jitter: Whether to add random jitter to delay.
-    """
-
-    model_config = ConfigDict(frozen=True, allow_inf_nan=False)
-
-    max_retries: int = Field(
-        default=3,
-        ge=0,
-        le=10,
-        description="Maximum number of retry attempts (0 disables retries)",
-    )
-    base_delay: float = Field(
-        default=1.0,
-        gt=0.0,
-        description="Initial delay in seconds before the first retry",
-    )
-    max_delay: float = Field(
-        default=60.0,
-        gt=0.0,
-        description="Upper bound on computed delay in seconds",
-    )
-    exponential_base: float = Field(
-        default=2.0,
-        gt=1.0,
-        description="Multiplier for exponential backoff",
-    )
-    jitter: bool = Field(
-        default=True,
-        description="Whether to add random jitter to delay",
-    )
-
-    @model_validator(mode="after")
-    def _validate_delay_ordering(self) -> Self:
-        """Ensure base_delay does not exceed max_delay."""
-        if self.base_delay > self.max_delay:
-            msg = (
-                f"base_delay ({self.base_delay}) must be"
-                f" <= max_delay ({self.max_delay})"
-            )
-            logger.warning(
-                CONFIG_VALIDATION_FAILED,
-                model="RetryConfig",
-                field="base_delay/max_delay",
-                base_delay=self.base_delay,
-                max_delay=self.max_delay,
-                reason=msg,
-            )
-            raise ValueError(msg)
-        return self
-
-
-class RateLimiterConfig(BaseModel):
-    """Configuration for client-side rate limiting.
-
-    Attributes:
-        max_requests_per_minute: Maximum requests per minute
-            (0 means unlimited).
-        max_concurrent: Maximum concurrent in-flight requests
-            (0 means unlimited).
-    """
-
-    model_config = ConfigDict(frozen=True, allow_inf_nan=False)
-
-    max_requests_per_minute: int = Field(
-        default=0,
-        ge=0,
-        description="Maximum requests per minute (0 = unlimited)",
-    )
-    max_concurrent: int = Field(
-        default=0,
-        ge=0,
-        description="Maximum concurrent in-flight requests (0 = unlimited)",
-    )
-
-
 class ProviderModelConfig(BaseModel):
     """Configuration for a single LLM model within a provider.
 
diff --git a/src/ai_company/core/resilience_config.py b/src/ai_company/core/resilience_config.py
new file mode 100644
index 0000000000..57685a79e5
--- /dev/null
+++ b/src/ai_company/core/resilience_config.py
@@ -0,0 +1,97 @@
+"""Resilience configuration models (retry + rate limiting).
+
+Defined in ``core/`` to avoid circular imports between ``config.schema``
+and ``providers.resilience``.  Both modules import from here.
+"""
+
+from typing import Self
+
+from pydantic import BaseModel, ConfigDict, Field, model_validator
+
+from ai_company.observability import get_logger
+from ai_company.observability.events.config import CONFIG_VALIDATION_FAILED
+
+logger = get_logger(__name__)
+
+
+class RetryConfig(BaseModel):
+    """Configuration for automatic retry of transient provider errors.
+
+    Attributes:
+        max_retries: Maximum number of retry attempts (0 disables retries).
+        base_delay: Initial delay in seconds before the first retry.
+        max_delay: Upper bound on computed delay in seconds.
+        exponential_base: Multiplier for exponential backoff.
+        jitter: Whether to add random jitter to delay.
+    """
+
+    model_config = ConfigDict(frozen=True, allow_inf_nan=False)
+
+    max_retries: int = Field(
+        default=3,
+        ge=0,
+        le=10,
+        description="Maximum number of retry attempts (0 disables retries)",
+    )
+    base_delay: float = Field(
+        default=1.0,
+        gt=0.0,
+        description="Initial delay in seconds before the first retry",
+    )
+    max_delay: float = Field(
+        default=60.0,
+        gt=0.0,
+        description="Upper bound on computed delay in seconds",
+    )
+    exponential_base: float = Field(
+        default=2.0,
+        gt=1.0,
+        description="Multiplier for exponential backoff",
+    )
+    jitter: bool = Field(
+        default=True,
+        description="Whether to add random jitter to delay",
+    )
+
+    @model_validator(mode="after")
+    def _validate_delay_ordering(self) -> Self:
+        """Ensure base_delay does not exceed max_delay."""
+        if self.base_delay > self.max_delay:
+            msg = (
+                f"base_delay ({self.base_delay}) must be"
+                f" <= max_delay ({self.max_delay})"
+            )
+            logger.warning(
+                CONFIG_VALIDATION_FAILED,
+                model="RetryConfig",
+                field="base_delay/max_delay",
+                base_delay=self.base_delay,
+                max_delay=self.max_delay,
+                reason=msg,
+            )
+            raise ValueError(msg)
+        return self
+
+
+class RateLimiterConfig(BaseModel):
+    """Configuration for client-side rate limiting.
+
+    Attributes:
+        max_requests_per_minute: Maximum requests per minute
+            (0 means unlimited).
+        max_concurrent: Maximum concurrent in-flight requests
+            (0 means unlimited).
+    """
+
+    model_config = ConfigDict(frozen=True, allow_inf_nan=False)
+
+    max_requests_per_minute: int = Field(
+        default=0,
+        ge=0,
+        description="Maximum requests per minute (0 = unlimited)",
+    )
+    max_concurrent: int = Field(
+        default=0,
+        ge=0,
+        description="Maximum concurrent in-flight requests (0 = unlimited)",
+    )
diff --git a/src/ai_company/engine/__init__.py b/src/ai_company/engine/__init__.py
index 138137888c..d9d43940c8 100644
--- a/src/ai_company/engine/__init__.py
+++ b/src/ai_company/engine/__init__.py
@@ -55,8 +55,6 @@
     TaskStructureClassifier,
 )
 from ai_company.engine.errors import (
-    BudgetExhaustedError,
-    DailyLimitExceededError,
     DecompositionCycleError,
     DecompositionDepthError,
     DecompositionError,
@@ -67,7 +65,6 @@
     NoEligibleAgentError,
     ParallelExecutionError,
     PromptBuildError,
-    QuotaExhaustedError,
     ResourceConflictError,
     TaskAssignmentError,
     TaskRoutingError,
@@ -171,12 +168,10 @@
     "AuctionAssignmentStrategy",
     "AutoTopologyConfig",
     "BudgetChecker",
-    "BudgetExhaustedError",
     "ClassificationResult",
     "CleanupCallback",
     "CooperativeTimeoutStrategy",
     "CostOptimizedAssignmentStrategy",
-    "DailyLimitExceededError",
     "DecompositionContext",
     "DecompositionCycleError",
     "DecompositionDepthError",
@@ -221,7 +216,6 @@
     "ProgressCallback",
     "PromptBuildError",
     "PromptTokenEstimator",
-    "QuotaExhaustedError",
     "ReactLoop",
     "RecoveryResult",
     "RecoveryStrategy",
diff --git a/src/ai_company/engine/agent_engine.py b/src/ai_company/engine/agent_engine.py
index 132b616fb4..4fe1a84382 100644
--- a/src/ai_company/engine/agent_engine.py
+++ b/src/ai_company/engine/agent_engine.py
@@ -9,6 +9,7 @@
 import time
 from typing import TYPE_CHECKING
 
+from ai_company.budget.errors import BudgetExhaustedError
 from ai_company.core.enums import TaskStatus
 from ai_company.engine._validation import (
     validate_agent,
@@ -18,10 +19,7 @@
 from ai_company.engine.classification.pipeline import classify_execution_errors
 from ai_company.engine.context import DEFAULT_MAX_TURNS, AgentContext
 from ai_company.engine.cost_recording import record_execution_costs
-from ai_company.engine.errors import (
-    BudgetExhaustedError,
-    ExecutionStateError,
-)
+from ai_company.engine.errors import ExecutionStateError
 from ai_company.engine.loop_protocol import (
     ExecutionResult,
     TerminationReason,
diff --git a/src/ai_company/engine/errors.py b/src/ai_company/engine/errors.py
index 42a1c4a636..7f5fdb7bce 100644
--- a/src/ai_company/engine/errors.py
+++ b/src/ai_company/engine/errors.py
@@ -21,31 +21,6 @@ class MaxTurnsExceededError(EngineError):
     """
 
 
-class BudgetExhaustedError(EngineError):
-    """Budget exhaustion signal for the engine layer.
-
-    Used in two contexts:
-
-    1. Raised directly by :meth:`BudgetEnforcer.check_can_execute`
-       when pre-flight budget checks fail (monthly hard stop or daily
-       limit exceeded).
-    2. Available for converting ``TerminationReason.BUDGET_EXHAUSTED``
-       loop results into a raised error at the engine layer.
-    """
-
-
-class DailyLimitExceededError(BudgetExhaustedError):
-    """Per-agent daily spending limit exceeded."""
-
-
-class QuotaExhaustedError(BudgetExhaustedError):
-    """Raised when provider quota is exhausted.
-
-    Currently raised for all degradation strategies. Degradation routing
-    (FALLBACK/QUEUE) is planned for a future milestone.
-    """
-
-
 class LoopExecutionError(EngineError):
     """Non-recoverable execution loop error for the engine layer.
 
diff --git a/src/ai_company/providers/__init__.py b/src/ai_company/providers/__init__.py
index 84c5952584..a198d4797d 100644
--- a/src/ai_company/providers/__init__.py
+++ b/src/ai_company/providers/__init__.py
@@ -38,8 +38,6 @@
 from .registry import ProviderRegistry
 from .resilience import (
     RateLimiter,
-    RateLimiterConfig,
-    RetryConfig,
     RetryExhaustedError,
     RetryHandler,
 )
@@ -107,9 +105,7 @@
     "ProviderTimeoutError",
     "RateLimitError",
     "RateLimiter",
-    "RateLimiterConfig",
     "ResolvedModel",
-    "RetryConfig",
     "RetryExhaustedError",
     "RetryHandler",
     "RoleBasedStrategy",
diff --git a/src/ai_company/providers/resilience/__init__.py b/src/ai_company/providers/resilience/__init__.py
index d8e6a5f1f1..442e3757de 100644
--- a/src/ai_company/providers/resilience/__init__.py
+++ b/src/ai_company/providers/resilience/__init__.py
@@ -1,18 +1,15 @@
 """Provider resilience infrastructure.
 
-Exports retry handling, rate limiting, configuration models,
-and the ``RetryExhaustedError`` for fallback-chain signaling.
+Exports retry handling, rate limiting, and the
+``RetryExhaustedError`` for fallback-chain signaling.
 """
 
-from .config import RateLimiterConfig, RetryConfig
 from .errors import RetryExhaustedError
 from .rate_limiter import RateLimiter
 from .retry import RetryHandler
 
 __all__ = [
     "RateLimiter",
-    "RateLimiterConfig",
-    "RetryConfig",
     "RetryExhaustedError",
     "RetryHandler",
 ]
diff --git a/src/ai_company/providers/resilience/config.py b/src/ai_company/providers/resilience/config.py
deleted file mode 100644
index bd9ac0ba2f..0000000000
--- a/src/ai_company/providers/resilience/config.py
+++ /dev/null
@@ -1,11 +0,0 @@
-"""Re-export resilience configuration models.
-
-Canonical definitions live in :mod:`ai_company.config.schema` to avoid
-circular imports (config → providers → config).  This module re-exports
-them so consumers can use ``from ai_company.providers.resilience.config
-import RetryConfig``.
-"""
-
-from ai_company.config.schema import RateLimiterConfig, RetryConfig
-
-__all__ = ["RateLimiterConfig", "RetryConfig"]
diff --git a/src/ai_company/providers/resilience/rate_limiter.py b/src/ai_company/providers/resilience/rate_limiter.py
index 31428fa70e..0f3035166a 100644
--- a/src/ai_company/providers/resilience/rate_limiter.py
+++ b/src/ai_company/providers/resilience/rate_limiter.py
@@ -4,14 +4,13 @@
 import math
 import time
 
+from ai_company.core.resilience_config import RateLimiterConfig  # noqa: TC001
 from ai_company.observability import get_logger
 from ai_company.observability.events.provider import (
     PROVIDER_RATE_LIMITER_PAUSED,
     PROVIDER_RATE_LIMITER_THROTTLED,
 )
 
-from .config import RateLimiterConfig  # noqa: TC001
-
 logger = get_logger(__name__)
 
 
diff --git a/src/ai_company/providers/resilience/retry.py b/src/ai_company/providers/resilience/retry.py
index b461b4649a..877e2ff274 100644
--- a/src/ai_company/providers/resilience/retry.py
+++ b/src/ai_company/providers/resilience/retry.py
@@ -18,7 +18,7 @@
 if TYPE_CHECKING:
     from collections.abc import Callable, Coroutine
 
-    from .config import RetryConfig
+    from ai_company.core.resilience_config import RetryConfig
 
 logger = get_logger(__name__)
 
diff --git a/src/ai_company/providers/routing/_strategy_helpers.py b/src/ai_company/providers/routing/_strategy_helpers.py
index 0d68be4e7a..ee82cbfead 100644
--- a/src/ai_company/providers/routing/_strategy_helpers.py
+++ b/src/ai_company/providers/routing/_strategy_helpers.py
@@ -4,7 +4,8 @@
 selection logic.
 """
 
-from ai_company.config.schema import RoutingConfig, RoutingRuleConfig  # noqa: TC001
+from typing import TYPE_CHECKING
+
 from ai_company.core.role_catalog import get_seniority_info
 from ai_company.observability import get_logger
 from ai_company.observability.events.routing import (
@@ -18,6 +19,9 @@
 from .models import ResolvedModel, RoutingDecision, RoutingRequest
 from .resolver import ModelResolver  # noqa: TC001
 
+if TYPE_CHECKING:
+    from ai_company.config.schema import RoutingConfig, RoutingRuleConfig
+
 logger = get_logger(__name__)
 
 
diff --git a/src/ai_company/providers/routing/resolver.py b/src/ai_company/providers/routing/resolver.py
index 76b6e264e9..0ba2250f8b 100644
--- a/src/ai_company/providers/routing/resolver.py
+++ b/src/ai_company/providers/routing/resolver.py
@@ -7,8 +7,8 @@
 """
 
 from types import MappingProxyType
+from typing import TYPE_CHECKING
 
-from ai_company.config.schema import ProviderConfig  # noqa: TC001
 from ai_company.observability import get_logger
 from ai_company.observability.events.routing import (
     ROUTING_MODEL_RESOLUTION_FAILED,
@@ -19,6 +19,9 @@
 from .errors import ModelResolutionError
 from .models import ResolvedModel
 
+if TYPE_CHECKING:
+    from ai_company.config.schema import ProviderConfig
+
 logger = get_logger(__name__)
 
 
diff --git a/src/ai_company/providers/routing/router.py b/src/ai_company/providers/routing/router.py
index 357c8ecbc8..bbf0b582ab 100644
--- a/src/ai_company/providers/routing/router.py
+++ b/src/ai_company/providers/routing/router.py
@@ -4,7 +4,8 @@
 Delegates to strategy implementations.
 """
 
-from ai_company.config.schema import ProviderConfig, RoutingConfig  # noqa: TC001
+from typing import TYPE_CHECKING
+
 from ai_company.observability import get_logger
 from ai_company.observability.events.routing import (
     ROUTING_DECISION_MADE,
@@ -18,6 +19,9 @@
 from .resolver import ModelResolver
 from .strategies import STRATEGY_MAP
 
+if TYPE_CHECKING:
+    from ai_company.config.schema import ProviderConfig, RoutingConfig
+
 logger = get_logger(__name__)
 
 
diff --git a/src/ai_company/providers/routing/strategies.py b/src/ai_company/providers/routing/strategies.py
index 8cd5255291..0077fc1eaf 100644
--- a/src/ai_company/providers/routing/strategies.py
+++ b/src/ai_company/providers/routing/strategies.py
@@ -7,9 +7,8 @@
 
 from collections.abc import Mapping  # noqa: TC003
 from types import MappingProxyType
-from typing import Final, NoReturn, Protocol, runtime_checkable
+from typing import TYPE_CHECKING, Final, NoReturn, Protocol, runtime_checkable
 
-from ai_company.config.schema import RoutingConfig  # noqa: TC001
 from ai_company.core.enums import SeniorityLevel  # noqa: TC001
 from ai_company.core.role_catalog import get_seniority_info
 from ai_company.observability import get_logger
@@ -35,6 +34,9 @@
 from .models import RoutingDecision, RoutingRequest
 from .resolver import ModelResolver  # noqa: TC001
 
+if TYPE_CHECKING:
+    from ai_company.config.schema import RoutingConfig
+
 logger = get_logger(__name__)
 
 # ── Strategy name constants ──────────────────────────────────────
diff --git a/tests/integration/providers/conftest.py b/tests/integration/providers/conftest.py
index 38a260fe5f..377904b4c1 100644
--- a/tests/integration/providers/conftest.py
+++ b/tests/integration/providers/conftest.py
@@ -13,11 +13,13 @@
 from litellm.types.utils import (  # type: ignore[attr-defined]
     ChatCompletionToolCallChunk,
     Delta,
+    ModelResponseStream,
     StreamingChoices,
     Usage,
 )
 
-from ai_company.config.schema import ProviderConfig, ProviderModelConfig, RetryConfig
+from ai_company.config.schema import ProviderConfig, ProviderModelConfig
+from ai_company.core.resilience_config import RetryConfig
 from ai_company.providers.enums import MessageRole
 from ai_company.providers.models import (
     ChatMessage,
@@ -165,9 +167,9 @@ def build_content_chunk(
     *,
     model: str = "test-model-001",
     chunk_id: str = "chunk_0",
-) -> ModelResponse:
+) -> ModelResponseStream:
     """Build a streaming chunk with text content."""
-    return ModelResponse(
+    return ModelResponseStream(
         id=chunk_id,
         choices=[
             StreamingChoices(
@@ -177,7 +179,6 @@ def build_content_chunk(
             ),
         ],
         model=model,
-        stream=True,
     )
 
 
@@ -187,9 +188,9 @@ def build_usage_chunk(
     completion_tokens: int = 50,
     model: str = "test-model-001",
     chunk_id: str = "chunk_usage",
-) -> ModelResponse:
+) -> ModelResponseStream:
     """Build a streaming chunk with usage data and no choices."""
-    return ModelResponse(
+    return ModelResponseStream(
         id=chunk_id,
         choices=[],
         usage=Usage(
@@ -198,7 +199,6 @@ def build_usage_chunk(
             total_tokens=prompt_tokens + completion_tokens,
         ),
         model=model,
-        stream=True,
     )
 
 
@@ -210,7 +210,7 @@ def build_tool_call_delta_chunk(  # noqa: PLR0913
     arguments: str | None = None,
     model: str = "test-model-001",
     chunk_id: str = "chunk_tc",
-) -> ModelResponse:
+) -> ModelResponseStream:
     """Build a streaming chunk with a tool call delta."""
     tc_delta = ChatCompletionToolCallChunk(
         index=index,
@@ -220,7 +220,7 @@ def build_tool_call_delta_chunk(  # noqa: PLR0913
         ),
         type="function",
     )
-    return ModelResponse(
+    return ModelResponseStream(
         id=chunk_id,
         choices=[
             StreamingChoices(
@@ -230,7 +230,6 @@ def build_tool_call_delta_chunk(  # noqa: PLR0913
             ),
         ],
         model=model,
-        stream=True,
     )
 
 
@@ -239,9 +238,9 @@ def build_finish_chunk(
     *,
     model: str = "test-model-001",
     chunk_id: str = "chunk_fin",
-) -> ModelResponse:
+) -> ModelResponseStream:
     """Build a streaming chunk with only a finish reason."""
-    return ModelResponse(
+    return ModelResponseStream(
         id=chunk_id,
         choices=[
             StreamingChoices(
@@ -251,14 +250,13 @@ def build_finish_chunk(
             ),
         ],
         model=model,
-        stream=True,
     )
 
 
 async def async_iter_chunks(
-    chunks: list[ModelResponse],
-) -> AsyncIterator[ModelResponse]:
-    """Wrap a list of ``ModelResponse`` chunks into an ``AsyncIterator``."""
+    chunks: list[ModelResponseStream],
+) -> AsyncIterator[ModelResponseStream]:
+    """Wrap a list of ``ModelResponseStream`` chunks into an ``AsyncIterator``."""
     for chunk in chunks:
         yield chunk
 
diff --git a/tests/integration/providers/test_retry_integration.py b/tests/integration/providers/test_retry_integration.py
index 642a5d3f24..2c675f8890 100644
--- a/tests/integration/providers/test_retry_integration.py
+++ b/tests/integration/providers/test_retry_integration.py
@@ -9,12 +9,8 @@
 
 import pytest
 
-from ai_company.config.schema import (
-    ProviderConfig,
-    ProviderModelConfig,
-    RateLimiterConfig,
-    RetryConfig,
-)
+from ai_company.config.schema import ProviderConfig, ProviderModelConfig
+from ai_company.core.resilience_config import RateLimiterConfig, RetryConfig
 from ai_company.providers.drivers.litellm_driver import LiteLLMDriver
 from ai_company.providers.enums import MessageRole
 from ai_company.providers.errors import (
diff --git a/tests/unit/budget/test_enforcer.py b/tests/unit/budget/test_enforcer.py
index 76c6283e43..ac29539a60 100644
--- a/tests/unit/budget/test_enforcer.py
+++ b/tests/unit/budget/test_enforcer.py
@@ -17,12 +17,12 @@
     BudgetConfig,
 )
 from ai_company.budget.enforcer import BudgetEnforcer
+from ai_company.budget.errors import BudgetExhaustedError, DailyLimitExceededError
 from ai_company.budget.tracker import CostTracker
 from ai_company.core.agent import AgentIdentity, ModelConfig
 from ai_company.core.enums import TaskStatus, TaskType
 from ai_company.core.task import Task
 from ai_company.engine.context import AgentContext
-from ai_company.engine.errors import BudgetExhaustedError, DailyLimitExceededError
 from ai_company.observability.events.budget import BUDGET_ALERT_THRESHOLD_CROSSED
 from ai_company.providers.models import TokenUsage
 from ai_company.providers.routing.models import ResolvedModel
diff --git a/tests/unit/budget/test_enforcer_quota.py b/tests/unit/budget/test_enforcer_quota.py
index 1df53c447a..f5eed304e1 100644
--- a/tests/unit/budget/test_enforcer_quota.py
+++ b/tests/unit/budget/test_enforcer_quota.py
@@ -11,6 +11,7 @@
 
 from ai_company.budget.config import BudgetAlertConfig, BudgetConfig
 from ai_company.budget.enforcer import BudgetEnforcer
+from ai_company.budget.errors import QuotaExhaustedError
 from ai_company.budget.quota import (
     QuotaCheckResult,
     QuotaLimit,
@@ -19,7 +20,6 @@
 )
 from ai_company.budget.quota_tracker import QuotaTracker
 from ai_company.budget.tracker import CostTracker
-from ai_company.engine.errors import QuotaExhaustedError
 
 pytestmark = pytest.mark.timeout(30)
 
diff --git a/tests/unit/budget/test_errors.py b/tests/unit/budget/test_errors.py
new file mode 100644
index 0000000000..499eb70af4
--- /dev/null
+++ b/tests/unit/budget/test_errors.py
@@ -0,0 +1,49 @@
+"""Tests for budget error hierarchy."""
+
+import pytest
+
+from ai_company.budget.errors import (
+    BudgetExhaustedError,
+    DailyLimitExceededError,
+    QuotaExhaustedError,
+)
+
+pytestmark = pytest.mark.timeout(30)
+
+
+@pytest.mark.unit
+class TestBudgetErrorHierarchy:
+    """Verify inheritance relationships in the budget error hierarchy."""
+
+    def test_budget_exhausted_is_exception(self) -> None:
+        assert issubclass(BudgetExhaustedError, Exception)
+
+    def test_daily_limit_is_budget_exhausted(self) -> None:
+        assert issubclass(DailyLimitExceededError, BudgetExhaustedError)
+        err = DailyLimitExceededError("daily limit hit")
+        assert isinstance(err, BudgetExhaustedError)
+
+    def test_quota_exhausted_is_budget_exhausted(self) -> None:
+        assert issubclass(QuotaExhaustedError, BudgetExhaustedError)
+        err = QuotaExhaustedError("quota hit")
+        assert isinstance(err, BudgetExhaustedError)
+
+    def test_budget_exhausted_not_engine_error(self) -> None:
+        """Budget errors are independent of the engine error hierarchy."""
+        from ai_company.engine.errors import EngineError
+
+        assert not issubclass(BudgetExhaustedError, EngineError)
+        assert not issubclass(DailyLimitExceededError, EngineError)
+        assert not issubclass(QuotaExhaustedError, EngineError)
+
+    def test_message_preserved(self) -> None:
+        msg = "agent-1 budget exhausted"
+        err = BudgetExhaustedError(msg)
+        assert str(err) == msg
+
+    def test_except_budget_exhausted_catches_subclasses(self) -> None:
+        """Ensure except BudgetExhaustedError catches all subtypes."""
+        for exc_cls in (DailyLimitExceededError, QuotaExhaustedError):
+            msg = "subclass caught"
+            with pytest.raises(BudgetExhaustedError):
+                raise exc_cls(msg)
diff --git a/tests/unit/config/conftest.py b/tests/unit/config/conftest.py
index 597d612cbf..f05e50b05d 100644
--- a/tests/unit/config/conftest.py
+++ b/tests/unit/config/conftest.py
@@ -14,14 +14,13 @@
     AgentConfig,
     ProviderConfig,
     ProviderModelConfig,
-    RateLimiterConfig,
-    RetryConfig,
     RootConfig,
     RoutingConfig,
     RoutingRuleConfig,
     TaskAssignmentConfig,
 )
 from ai_company.core.company import CompanyConfig
+from ai_company.core.resilience_config import RateLimiterConfig, RetryConfig
 from ai_company.hr.promotion.config import PromotionConfig
 from ai_company.memory.config import CompanyMemoryConfig
 from ai_company.memory.org.config import OrgMemoryConfig
diff --git a/tests/unit/engine/test_agent_engine_budget.py b/tests/unit/engine/test_agent_engine_budget.py
index 4c095a480c..324e2b78bb 100644
--- a/tests/unit/engine/test_agent_engine_budget.py
+++ b/tests/unit/engine/test_agent_engine_budget.py
@@ -10,9 +10,13 @@
     BudgetConfig,
 )
 from ai_company.budget.enforcer import BudgetEnforcer
+from ai_company.budget.errors import (
+    BudgetExhaustedError,
+    DailyLimitExceededError,
+    QuotaExhaustedError,
+)
 from ai_company.budget.tracker import CostTracker
 from ai_company.engine.agent_engine import AgentEngine
-from ai_company.engine.errors import BudgetExhaustedError, DailyLimitExceededError
 from ai_company.engine.loop_protocol import TerminationReason
 
 if TYPE_CHECKING:
@@ -51,8 +55,9 @@ class TestEngineWithEnforcer:
         [
             (BudgetExhaustedError, "Monthly budget exhausted"),
             (DailyLimitExceededError, "Daily limit exceeded"),
+            (QuotaExhaustedError, "Provider quota exhausted"),
         ],
-        ids=["monthly_exhausted", "daily_limit"],
+        ids=["monthly_exhausted", "daily_limit", "quota_exhausted"],
     )
     async def test_preflight_budget_stop_returns_budget_exhausted(
         self,
diff --git a/tests/unit/engine/test_errors.py b/tests/unit/engine/test_errors.py
index d98346b4ce..38ba7ea75a 100644
--- a/tests/unit/engine/test_errors.py
+++ b/tests/unit/engine/test_errors.py
@@ -2,8 +2,8 @@
 
 import pytest
 
+from ai_company.budget.errors import BudgetExhaustedError
 from ai_company.engine.errors import (
-    BudgetExhaustedError,
     EngineError,
     ExecutionStateError,
     LoopExecutionError,
@@ -34,10 +34,10 @@ def test_prompt_build_error_is_engine_error(self) -> None:
         err = PromptBuildError("test")
         assert isinstance(err, EngineError)
 
-    def test_budget_exhausted_error_is_engine_error(self) -> None:
-        assert issubclass(BudgetExhaustedError, EngineError)
+    def test_budget_exhausted_error_is_not_engine_error(self) -> None:
+        assert not issubclass(BudgetExhaustedError, EngineError)
         err = BudgetExhaustedError("out of budget")
-        assert isinstance(err, EngineError)
+        assert isinstance(err, Exception)
         assert str(err) == "out of budget"
 
     def test_loop_execution_error_is_engine_error(self) -> None:
diff --git a/tests/unit/providers/drivers/conftest.py b/tests/unit/providers/drivers/conftest.py
index ecd89d22a7..fba9d14d85 100644
--- a/tests/unit/providers/drivers/conftest.py
+++ b/tests/unit/providers/drivers/conftest.py
@@ -5,12 +5,8 @@
 
 import pytest
 
-from ai_company.config.schema import (
-    ProviderConfig,
-    ProviderModelConfig,
-    RateLimiterConfig,
-    RetryConfig,
-)
+from ai_company.config.schema import ProviderConfig, ProviderModelConfig
+from ai_company.core.resilience_config import RateLimiterConfig, RetryConfig
 
 if TYPE_CHECKING:
     from collections.abc import AsyncIterator
diff --git a/tests/unit/providers/resilience/conftest.py b/tests/unit/providers/resilience/conftest.py
index 6adaf59c72..ec6e20f71e 100644
--- a/tests/unit/providers/resilience/conftest.py
+++ b/tests/unit/providers/resilience/conftest.py
@@ -2,6 +2,7 @@
 
 import pytest
 
+from ai_company.core.resilience_config import RateLimiterConfig, RetryConfig
 from ai_company.providers.errors import (
     AuthenticationError,
     ProviderConnectionError,
@@ -9,7 +10,6 @@
     ProviderTimeoutError,
     RateLimitError,
 )
-from ai_company.providers.resilience.config import RateLimiterConfig, RetryConfig
 
 
 @pytest.fixture
diff --git a/tests/unit/providers/resilience/test_config.py b/tests/unit/providers/resilience/test_config.py
index 98b2381cf4..795f497680 100644
--- a/tests/unit/providers/resilience/test_config.py
+++ b/tests/unit/providers/resilience/test_config.py
@@ -3,7 +3,7 @@
 import pytest
 from pydantic import ValidationError
 
-from ai_company.providers.resilience.config import RateLimiterConfig, RetryConfig
+from ai_company.core.resilience_config import RateLimiterConfig, RetryConfig
 
 pytestmark = pytest.mark.timeout(30)
 
diff --git a/tests/unit/providers/resilience/test_rate_limiter.py b/tests/unit/providers/resilience/test_rate_limiter.py
index ef52120585..e13ef23e38 100644
--- a/tests/unit/providers/resilience/test_rate_limiter.py
+++ b/tests/unit/providers/resilience/test_rate_limiter.py
@@ -7,11 +7,11 @@
 import pytest
 import structlog
 
+from ai_company.core.resilience_config import RateLimiterConfig
 from ai_company.observability.events.provider import (
     PROVIDER_RATE_LIMITER_PAUSED,
     PROVIDER_RATE_LIMITER_THROTTLED,
 )
-from ai_company.providers.resilience.config import RateLimiterConfig
 from ai_company.providers.resilience.rate_limiter import RateLimiter
 
 pytestmark = pytest.mark.timeout(30)
diff --git a/tests/unit/providers/resilience/test_retry.py b/tests/unit/providers/resilience/test_retry.py
index 80c6181a21..98b0a05c24 100644
--- a/tests/unit/providers/resilience/test_retry.py
+++ b/tests/unit/providers/resilience/test_retry.py
@@ -5,6 +5,7 @@
 import pytest
 import structlog
 
+from ai_company.core.resilience_config import RetryConfig
 from ai_company.observability.events.provider import (
     PROVIDER_CALL_ERROR,
     PROVIDER_RETRY_ATTEMPT,
@@ -18,7 +19,6 @@
     ProviderTimeoutError,
     RateLimitError,
 )
-from ai_company.providers.resilience.config import RetryConfig
 from ai_company.providers.resilience.errors import RetryExhaustedError
 from ai_company.providers.resilience.retry import RetryHandler
 
diff --git a/uv.lock b/uv.lock
index 2805dc3cd7..d12431845f 100644
--- a/uv.lock
+++ b/uv.lock
@@ -52,7 +52,7 @@ requires-dist = [
     { name = "aiosqlite", specifier = "==0.22.1" },
     { name = "jinja2", specifier = "==3.1.6" },
     { name = "jsonschema", specifier = "==4.26.0" },
-    { name = "litellm", specifier = "==1.82.0" },
+    { name = "litellm", specifier = "==1.82.1" },
     { name = "litestar", extras = ["brotli", "prometheus", "pydantic", "standard", "structlog"], specifier = "==2.21.1" },
     { name = "mcp", specifier = "==1.26.0" },
     { name = "pydantic", specifier = "==2.12.5" },
@@ -893,7 +893,7 @@ wheels = [
 
 [[package]]
 name = "litellm"
-version = "1.82.0"
+version = "1.82.1"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "aiohttp" },
@@ -909,9 +909,9 @@ dependencies = [
     { name = "tiktoken" },
     { name = "tokenizers" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/6c/00/49bb5c28e0dea0f5086229a2a08d5fdc6c8dc0d8e2acb2a2d1f7dd9f4b70/litellm-1.82.0.tar.gz", hash = "sha256:d388f52447daccbcaafa19a3e68d17b75f1374b5bf2cde680d65e1cd86e50d22", size = 16800355, upload-time = "2026-03-01T02:35:30.363Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/34/bd/6251e9a965ae2d7bc3342ae6c1a2d25dd265d354c502e63225451b135016/litellm-1.82.1.tar.gz", hash = "sha256:bc8427cdccc99e191e08e36fcd631c93b27328d1af789839eb3ac01a7d281890", size = 17197496, upload-time = "2026-03-10T09:10:04.438Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/28/89/eb28bfcf97d6b045c400e72eb047c381594467048c237dbb6c227764084c/litellm-1.82.0-py3-none-any.whl", hash = "sha256:5496b5d4532cccdc7a095c21cbac4042f7662021c57bc1d17be4e39838929e80", size = 14911978, upload-time = "2026-03-01T02:35:26.844Z" },
+    { url = "https://files.pythonhosted.org/packages/57/77/0c6eca2cb049793ddf8ce9cdcd5123a35666c4962514788c4fc90edf1d3b/litellm-1.82.1-py3-none-any.whl", hash = "sha256:a9ec3fe42eccb1611883caaf8b1bf33c9f4e12163f94c7d1004095b14c379eb2", size = 15341896, upload-time = "2026-03-10T09:10:00.702Z" },
 ]
 
 [[package]]