truthound-dashboard 1.4.4__py3-none-any.whl → 1.5.0__py3-none-any.whl

truthound_dashboard/db/repository.py

@@ -188,6 +188,8 @@ class BaseRepository(Generic[ModelT]):
         if instance is None:
             return False
 
+        # Refresh to load all relationships for proper cascade delete
+        await self.session.refresh(instance)
         await self.session.delete(instance)
         await self.session.flush()
         return True

truthound_dashboard/main.py

@@ -108,6 +108,13 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     await ws_manager.start()
     logger.info("WebSocket manager started")
 
+    # Start streaming session cleanup
+    from truthound_dashboard.core.streaming_anomaly import get_streaming_detector
+
+    streaming_detector = get_streaming_detector()
+    await streaming_detector.start()
+    logger.info("Streaming session cleanup started")
+
     # Register maintenance tasks with scheduler
     _register_maintenance_tasks()
 
@@ -116,6 +123,10 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # Shutdown
     logger.info("Shutting down Truthound Dashboard")
 
+    # Stop streaming session cleanup
+    await streaming_detector.stop()
+    logger.info("Streaming session cleanup stopped")
+
     # Stop WebSocket manager
     await ws_manager.stop()
     logger.info("WebSocket manager stopped")

truthound_dashboard/schemas/__init__.py

@@ -20,8 +20,6 @@ from .base import (
 )
 from .drift import (
     ColumnDriftResult,
-    CorrectionMethod,
-    CorrectionMethodLiteral,
     DEFAULT_THRESHOLDS,
     DriftCompareRequest,
     DriftComparisonListItem,
@@ -45,14 +43,9 @@ from .profile import (
     ColumnProfile,
     DetectedPattern,
     HistogramBucket,
-    PatternDetectionConfig,
-    PatternType,
+    ProfileAdvancedRequest,
     ProfileRequest,
     ProfileResponse,
-    SamplingConfig,
-    SamplingMetadata,
-    SamplingStrategy,
-    SamplingStrategyType,
 )
 from .rule import (
     RuleBase,
@@ -375,6 +368,97 @@ from .plugins import (
     ValidatorTestRequest,
     ValidatorTestResponse,
 )
+# Storage Tiering (truthound 1.2.10+)
+from .tiering import (
+    AccessBasedPolicyConfig,
+    AgeBasedPolicyConfig,
+    CompositePolicyConfig,
+    CustomPolicyConfig,
+    MigrationDirection,
+    MigrationHistoryCreate,
+    MigrationHistoryListResponse,
+    MigrationHistoryResponse,
+    MigrationStatus,
+    PolicyConfigBase,
+    PolicyTypeInfo,
+    PolicyTypesResponse,
+    ScheduledPolicyConfig,
+    SizeBasedPolicyConfig,
+    StorageTierBase,
+    StorageTierCreate,
+    StorageTierListResponse,
+    StorageTierResponse,
+    StorageTierUpdate,
+    TierPolicyBase,
+    TierPolicyCreate,
+    TierPolicyListResponse,
+    TierPolicyResponse,
+    TierPolicyType,
+    TierPolicyUpdate,
+    TierPolicyWithChildren,
+    TierStatistics,
+    TierType,
+    TieringConfigBase,
+    TieringConfigCreate,
+    TieringConfigListResponse,
+    TieringConfigResponse,
+    TieringConfigUpdate,
+    TieringStatistics,
+)
+# Quality Reporter
+from .quality_reporter import (
+    ChartType,
+    ConfusionMatrixSchema,
+    QualityCompareRequest,
+    QualityCompareResponse,
+    QualityDisplayMode,
+    QualityFilterRequest,
+    QualityFormatsResponse,
+    QualityLevel,
+    QualityLevelDistribution,
+    QualityMetricsSchema,
+    QualityReportConfigSchema,
+    QualityReportFormat,
+    QualityReportGenerateRequest,
+    QualityReportListResponse,
+    QualityReportResponse,
+    QualityReportStatus,
+    QualityScoreListResponse,
+    QualityScoreRequest,
+    QualityScoreResponse,
+    QualityScoreSchema,
+    QualityStatisticsSchema,
+    QualitySummaryResponse,
+    QualityThresholdsSchema,
+    ReportSortOrder,
+)
+# Observability (truthound store observability)
+from .observability import (
+    AuditEventListResponse,
+    AuditEventResponse,
+    AuditEventTypeEnum,
+    AuditQueryRequest,
+    AuditStatsResponse,
+    AuditStatusEnum,
+    HistogramValue,
+    MetricsResponse,
+    MetricTypeEnum,
+    MetricValue,
+    ObservabilityConfigRequest,
+    ObservabilityConfigResponse,
+    ObservabilityStatsResponse,
+    SpanContext,
+    SpanEvent,
+    SpanKindEnum,
+    SpanListResponse,
+    SpanResponse,
+    SpanStatusEnum,
+    StoreMetricsResponse,
+    SummaryQuantile,
+    SummaryValue,
+    TraceResponse,
+    TracingStatsResponse,
+)
 
 __all__ = [
     # Base
@@ -432,12 +516,6 @@ __all__ = [
     "ProfileResponse",
     "DetectedPattern",
     "HistogramBucket",
-    "PatternDetectionConfig",
-    "PatternType",
-    "SamplingConfig",
-    "SamplingMetadata",
-    "SamplingStrategy",
-    "SamplingStrategyType",
     # Data Masking
     "MaskingStrategy",
     "MaskingStrategyLiteral",
@@ -466,8 +544,6 @@ __all__ = [
     # Drift
     "DriftMethod",
     "DriftMethodLiteral",
-    "CorrectionMethod",
-    "CorrectionMethodLiteral",
     "DEFAULT_THRESHOLDS",
     "get_default_threshold",
     "DriftCompareRequest",
@@ -707,4 +783,89 @@ __all__ = [
     "ReporterGenerateResponse",
     "MarketplaceSearchRequest",
     "MarketplaceStats",
+    # Storage Tiering (truthound 1.2.10+)
+    "TierType",
+    "MigrationDirection",
+    "TierPolicyType",
+    "MigrationStatus",
+    "StorageTierBase",
+    "StorageTierCreate",
+    "StorageTierUpdate",
+    "StorageTierResponse",
+    "StorageTierListResponse",
+    "PolicyConfigBase",
+    "AgeBasedPolicyConfig",
+    "AccessBasedPolicyConfig",
+    "SizeBasedPolicyConfig",
+    "ScheduledPolicyConfig",
+    "CompositePolicyConfig",
+    "CustomPolicyConfig",
+    "TierPolicyBase",
+    "TierPolicyCreate",
+    "TierPolicyUpdate",
+    "TierPolicyResponse",
+    "TierPolicyListResponse",
+    "TierPolicyWithChildren",
+    "TieringConfigBase",
+    "TieringConfigCreate",
+    "TieringConfigUpdate",
+    "TieringConfigResponse",
+    "TieringConfigListResponse",
+    "MigrationHistoryCreate",
+    "MigrationHistoryResponse",
+    "MigrationHistoryListResponse",
+    "TierStatistics",
+    "TieringStatistics",
+    "PolicyTypeInfo",
+    "PolicyTypesResponse",
+    # Quality Reporter
+    "QualityLevel",
+    "QualityReportFormat",
+    "QualityReportStatus",
+    "QualityDisplayMode",
+    "ReportSortOrder",
+    "ChartType",
+    "ConfusionMatrixSchema",
+    "QualityMetricsSchema",
+    "QualityThresholdsSchema",
+    "QualityScoreSchema",
+    "QualityStatisticsSchema",
+    "QualityLevelDistribution",
+    "QualityScoreRequest",
+    "QualityFilterRequest",
+    "QualityReportConfigSchema",
+    "QualityReportGenerateRequest",
+    "QualityCompareRequest",
+    "QualityScoreResponse",
+    "QualityReportResponse",
+    "QualityCompareResponse",
+    "QualitySummaryResponse",
+    "QualityScoreListResponse",
+    "QualityReportListResponse",
+    "QualityFormatsResponse",
+    # Observability (truthound store observability)
+    "AuditEventTypeEnum",
+    "AuditStatusEnum",
+    "MetricTypeEnum",
+    "SpanKindEnum",
+    "SpanStatusEnum",
+    "ObservabilityConfigRequest",
+    "ObservabilityConfigResponse",
+    "AuditEventResponse",
+    "AuditEventListResponse",
+    "AuditQueryRequest",
+    "AuditStatsResponse",
+    "MetricValue",
+    "HistogramValue",
+    "SummaryQuantile",
+    "SummaryValue",
+    "MetricsResponse",
+    "StoreMetricsResponse",
+    "SpanContext",
+    "SpanEvent",
+    "SpanResponse",
+    "SpanListResponse",
+    "TraceResponse",
+    "TracingStatsResponse",
+    "ObservabilityStatsResponse",
 ]

truthound_dashboard/schemas/base.py

@@ -3,12 +3,18 @@
 This module provides reusable base classes and mixins for Pydantic schemas,
 enabling consistent patterns across all API schemas.
 
+Design Philosophy - Direct Response Style:
+- Single resources return the resource directly (no wrapper)
+- List endpoints return PaginatedResponse[T] with data, total, offset, limit
+- Errors are handled via HTTPException (FastAPI converts to proper JSON)
+- Success is indicated by HTTP status codes (200, 201, 204), not response fields
+
 The schema classes follow a consistent naming convention:
 - *Base: Common fields shared by create/update/response
 - *Create: Fields for creation (POST)
 - *Update: Fields for updates (PUT/PATCH)
 - *Response: Fields returned in responses
-- *ListResponse: Paginated list response wrapper
+- *ListResponse: Alias for PaginatedResponse[*Response]
 """
 
 from __future__ import annotations
@@ -49,28 +55,20 @@ class IDMixin:
     id: str = Field(..., description="Unique identifier")
 
 
-class ResponseWrapper(BaseSchema, Generic[T]):
-    """Generic wrapper for single item responses.
-
-    Provides consistent structure for API responses.
-    """
-
-    success: bool = Field(default=True, description="Whether request succeeded")
-    data: T = Field(..., description="Response data")
-    message: str | None = Field(default=None, description="Optional message")
+class PaginatedResponse(BaseSchema, Generic[T]):
+    """Generic paginated list response.
 
+    RESTful design - no 'success' field. HTTP status codes indicate success/failure.
 
-# Alias for backward compatibility with newer code
-DataResponse = ResponseWrapper
-
-
-class ListResponseWrapper(BaseSchema, Generic[T]):
-    """Generic wrapper for list responses with pagination.
-
-    Provides consistent structure for paginated API responses.
+    Example response:
+        {
+            "data": [...],
+            "total": 100,
+            "offset": 0,
+            "limit": 20
+        }
     """
 
-    success: bool = Field(default=True, description="Whether request succeeded")
     data: list[T] = Field(default_factory=list, description="List of items")
     total: int = Field(default=0, description="Total count of items")
     offset: int = Field(default=0, description="Offset for pagination")
@@ -82,10 +80,34 @@ class ListResponseWrapper(BaseSchema, Generic[T]):
         return self.offset + len(self.data) < self.total
 
 
+# Backward compatibility aliases
+ListResponseWrapper = PaginatedResponse
+"""@deprecated Use PaginatedResponse instead."""
+
+
+class ResponseWrapper(BaseSchema, Generic[T]):
+    """@deprecated - For backward compatibility only.
+
+    New code should return resources directly without wrapper.
+    """
+
+    success: bool = Field(default=True, description="Whether request succeeded")
+    data: T = Field(..., description="Response data")
+    message: str | None = Field(default=None, description="Optional message")
+
+
+# Alias for backward compatibility
+DataResponse = ResponseWrapper
+"""@deprecated Use direct response instead."""
+
+
 class ErrorResponse(BaseSchema):
-    """Standard error response schema."""
+    """Standard error response schema.
+
+    Note: In most cases, use HTTPException instead. This schema is for
+    documentation purposes and complex error responses.
+    """
 
-    success: bool = Field(default=False)
     detail: str = Field(..., description="Error description")
     code: str | None = Field(default=None, description="Error code")
     errors: list[dict[str, Any]] | None = Field(
@@ -94,7 +116,6 @@ class ErrorResponse(BaseSchema):
 
 
 class MessageResponse(BaseSchema):
-    """Simple message response schema."""
+    """Simple message response schema for delete/action endpoints."""
 
-    success: bool = Field(default=True)
     message: str = Field(..., description="Response message")

truthound_dashboard/schemas/collaboration.py

@@ -89,11 +89,24 @@ class CommentResponse(BaseSchema, IDMixin, TimestampMixin):
     def from_model(cls, comment: any, include_replies: bool = True) -> CommentResponse:
         """Create response from model."""
         replies = []
-        if include_replies and comment.replies:
-            replies = [
-                CommentResponse.from_model(r, include_replies=False)
-                for r in comment.replies
-            ]
+        reply_count = 0
+
+        if include_replies:
+            # Access replies if loaded (include_replies=True means we're a parent)
+            try:
+                loaded_replies = comment.replies
+                replies = [
+                    CommentResponse.from_model(r, include_replies=False)
+                    for r in loaded_replies
+                ]
+                reply_count = len(loaded_replies)
+            except Exception:
+                # If replies not loaded, use 0
+                reply_count = 0
+        else:
+            # For nested replies, avoid lazy loading - just use 0
+            # Nested reply counts aren't typically needed in UI
+            reply_count = 0
 
         return cls(
             id=comment.id,
@@ -103,7 +116,7 @@ class CommentResponse(BaseSchema, IDMixin, TimestampMixin):
             author_id=comment.author_id,
             parent_id=comment.parent_id,
             is_reply=comment.is_reply,
-            reply_count=comment.reply_count,
+            reply_count=reply_count,
             replies=replies,
             created_at=comment.created_at,
             updated_at=comment.updated_at,

truthound_dashboard/schemas/cross_alerts.py

@@ -2,6 +2,12 @@
 
 This module defines schemas for cross-feature integration between
 Anomaly Detection and Drift Monitoring alerts.
+
+API Design: Direct Response Style
+- Single resources return the resource directly
+- List endpoints return PaginatedResponse with data, total, offset, limit
+- Errors are handled via HTTPException
+- Success is indicated by HTTP status codes (200, 201, 204)
 """
 
 from __future__ import annotations
@@ -11,7 +17,7 @@ from typing import Literal
 
 from pydantic import Field
 
-from .base import BaseSchema, IDMixin, TimestampMixin, ListResponseWrapper
+from .base import BaseSchema, IDMixin, TimestampMixin, PaginatedResponse
 
 
 # =============================================================================
@@ -105,12 +111,22 @@ class CrossAlertCorrelation(IDMixin, TimestampMixin, BaseSchema):
     )
 
 
-class CrossAlertCorrelationListResponse(ListResponseWrapper[CrossAlertCorrelation]):
+class CrossAlertCorrelationListResponse(PaginatedResponse[CrossAlertCorrelation]):
     """Paginated list of cross-alert correlations."""
 
     pass
 
 
+class CorrelationSearchResult(BaseSchema):
+    """Result of a correlation search for a specific source."""
+
+    correlations: list[CrossAlertCorrelation] = Field(
+        default_factory=list,
+        description="Found correlations",
+    )
+    total: int = Field(default=0, description="Total correlations found")
+
+
 # =============================================================================
 # Auto-Trigger Configuration Schemas
 # =============================================================================
@@ -272,7 +288,7 @@ class AutoTriggerEvent(IDMixin, TimestampMixin, BaseSchema):
     )
 
 
-class AutoTriggerEventListResponse(ListResponseWrapper[AutoTriggerEvent]):
+class AutoTriggerEventListResponse(PaginatedResponse[AutoTriggerEvent]):
     """Paginated list of auto-trigger events."""
 
     pass

truthound_dashboard/schemas/drift.py

@@ -2,20 +2,27 @@
 
 Schemas for drift comparison request/response.
 
-Drift Methods (from truthound):
-- ks: Kolmogorov-Smirnov test (continuous distributions)
-- psi: Population Stability Index (any distribution, industry standard)
-- chi2: Chi-Square test (categorical data)
-- js: Jensen-Shannon divergence (probability distributions)
-- kl: Kullback-Leibler divergence (distribution difference)
-- wasserstein: Wasserstein/Earth Mover's Distance (distribution transport)
-- cvm: Cramér-von Mises test (more sensitive to tails than KS)
-- anderson: Anderson-Darling test (weighted for tail sensitivity)
-
-Multiple Testing Correction:
-- bonferroni: Conservative, independent tests
-- holm: Sequential adjustment, less conservative
-- bh: Benjamini-Hochberg (FDR control, default for multiple columns)
+Drift Methods (14 supported by th.compare() in truthound v1.2.9+):
+
+General purpose (any column type):
+- auto: Smart selection (numeric → PSI, categorical → chi2) [RECOMMENDED]
+- js: Jensen-Shannon divergence (symmetric, bounded 0-1)
+- hellinger: Hellinger distance (bounded metric, 0-1)
+- bhattacharyya: Bhattacharyya distance (classification error bounds)
+- tv: Total Variation distance (max probability difference)
+
+Numeric columns only:
+- ks: Kolmogorov-Smirnov test
+- psi: Population Stability Index (industry standard)
+- kl: Kullback-Leibler divergence (asymmetric)
+- wasserstein: Wasserstein/Earth Mover's Distance
+- cvm: Cramér-von Mises test (tail sensitive)
+- anderson: Anderson-Darling test (most tail sensitive)
+- energy: Energy distance (location/scale sensitive)
+- mmd: Maximum Mean Discrepancy (kernel-based, high-dimensional)
+
+Categorical columns:
+- chi2: Chi-Square test
 """
 
 from __future__ import annotations
@@ -29,18 +36,29 @@ from .base import IDMixin, TimestampMixin
 
 
 class DriftMethod(str, Enum):
-    """Drift detection methods supported by truthound.
+    """Drift detection methods supported by truthound v1.2.9+.
 
-    Each method has different characteristics and use cases:
+    All 14 methods are fully supported by th.compare():
+
+    General purpose (any column type):
     - auto: Smart selection based on data type (numeric → PSI, categorical → chi2)
-    - ks: Kolmogorov-Smirnov test - best for continuous distributions
-    - psi: Population Stability Index - industry standard, any distribution
-    - chi2: Chi-Square test - best for categorical data
     - js: Jensen-Shannon divergence - symmetric, bounded (0-1)
-    - kl: Kullback-Leibler divergence - information loss measure
-    - wasserstein: Earth Mover's Distance - metric, meaningful for non-overlapping
-    - cvm: Cramér-von Mises - more sensitive to tail differences than KS
-    - anderson: Anderson-Darling - weighted for tail sensitivity
+    - hellinger: Hellinger distance - bounded metric (0-1), triangle inequality
+    - bhattacharyya: Bhattacharyya distance - classification error bounds
+    - tv: Total Variation distance - max probability difference
+
+    Numeric columns only:
+    - ks: Kolmogorov-Smirnov test - detects any distribution shape difference
+    - psi: Population Stability Index - industry standard for model monitoring
+    - kl: Kullback-Leibler divergence - measures information loss (asymmetric)
+    - wasserstein: Earth Mover's Distance - intuitive physical interpretation
+    - cvm: Cramér-von Mises - more sensitive to tails than KS
+    - anderson: Anderson-Darling - most sensitive to tail differences
+    - energy: Energy distance - location/scale sensitive
+    - mmd: Maximum Mean Discrepancy - kernel-based, high-dimensional
+
+    Categorical columns:
+    - chi2: Chi-Square test - best for categorical data
     """
 
     AUTO = "auto"
@@ -52,23 +70,12 @@ class DriftMethod(str, Enum):
     WASSERSTEIN = "wasserstein"
     CVM = "cvm"
     ANDERSON = "anderson"
-
-
-class CorrectionMethod(str, Enum):
-    """Multiple testing correction methods.
-
-    When comparing multiple columns, correction adjusts p-values to control
-    false discovery rate:
-    - none: No correction (use with caution)
-    - bonferroni: Conservative, suitable for independent tests
-    - holm: Sequential adjustment, less conservative than Bonferroni
-    - bh: Benjamini-Hochberg (FDR control), default for multiple columns
-    """
-
-    NONE = "none"
-    BONFERRONI = "bonferroni"
-    HOLM = "holm"
-    BH = "bh"
+    # New in v1.2.9
+    HELLINGER = "hellinger"
+    BHATTACHARYYA = "bhattacharyya"
+    TV = "tv"
+    ENERGY = "energy"
+    MMD = "mmd"
 
 
 # Default thresholds for each detection method
@@ -82,6 +89,12 @@ DEFAULT_THRESHOLDS: dict[DriftMethod, float] = {
     DriftMethod.WASSERSTEIN: 0.1,  # Scale-dependent, adjust based on data
     DriftMethod.CVM: 0.05,
     DriftMethod.ANDERSON: 0.05,
+    # New in v1.2.9
+    DriftMethod.HELLINGER: 0.1,
+    DriftMethod.BHATTACHARYYA: 0.1,
+    DriftMethod.TV: 0.1,
+    DriftMethod.ENERGY: 0.1,
+    DriftMethod.MMD: 0.1,
 }
 
 
@@ -104,9 +117,9 @@ def get_default_threshold(method: DriftMethod | str) -> float:
 
 # Type alias for method values (for Literal type hints)
 DriftMethodLiteral = Literal[
-    "auto", "ks", "psi", "chi2", "js", "kl", "wasserstein", "cvm", "anderson"
+    "auto", "ks", "psi", "chi2", "js", "kl", "wasserstein", "cvm", "anderson",
+    "hellinger", "bhattacharyya", "tv", "energy", "mmd"
 ]
-CorrectionMethodLiteral = Literal["none", "bonferroni", "holm", "bh"]
 
 
 class DriftCompareRequest(BaseModel):
@@ -120,10 +133,9 @@ class DriftCompareRequest(BaseModel):
     method: DriftMethodLiteral = Field(
         "auto",
         description=(
-            "Drift detection method: "
-            "auto (smart selection), ks (Kolmogorov-Smirnov), psi (Population Stability Index), "
-            "chi2 (Chi-Square), js (Jensen-Shannon), kl (Kullback-Leibler), "
-            "wasserstein (Earth Mover's), cvm (Cramér-von Mises), anderson (Anderson-Darling)"
+            "Drift detection method (14 available): "
+            "auto (smart selection), ks, psi, chi2, js, kl, wasserstein, cvm, anderson, "
+            "hellinger, bhattacharyya, tv, energy, mmd"
         ),
     )
     threshold: float | None = Field(
@@ -132,13 +144,6 @@ class DriftCompareRequest(BaseModel):
         le=1,
         description="Custom threshold (default varies by method: KS/chi2/cvm/anderson=0.05, PSI/JS/KL/wasserstein=0.1)",
     )
-    correction: CorrectionMethodLiteral | None = Field(
-        None,
-        description=(
-            "Multiple testing correction: none, bonferroni (conservative), "
-            "holm (sequential), bh (Benjamini-Hochberg FDR, default for multiple columns)"
-        ),
-    )
     sample_size: int | None = Field(
         None, ge=100, description="Sample size for large datasets"
     )
@@ -226,8 +231,9 @@ class DriftComparisonListItem(BaseModel, IDMixin, TimestampMixin):
 
 
 class DriftComparisonListResponse(BaseModel):
-    """List response for drift comparisons."""
+    """Paginated list response for drift comparisons."""
 
-    success: bool = True
     data: list[DriftComparisonListItem] = Field(default_factory=list)
     total: int = 0
+    offset: int = 0
+    limit: int = 100