omnibase_infra 0.2.1__py3-none-any.whl → 0.2.2__py3-none-any.whl

omnibase_infra/runtime/transition_notification_publisher.py

@@ -0,0 +1,764 @@
+# SPDX-License-Identifier: MIT
+# Copyright (c) 2025 OmniNode Team
+"""
+Transition Notification Publisher Implementation.
+
+Publishes state transition notifications after projection commits. This enables
+orchestrators to reliably detect state transitions via the Observer pattern,
+maintaining loose coupling between reducers and workflow coordinators.
+
+Architecture Overview:
+    This service implements post-commit notification publishing in the ONEX
+    state machine architecture:
+
+    1. Reducers commit state transitions to projections
+    2. Post-commit hook creates ModelStateTransitionNotification
+    3. TransitionNotificationPublisher publishes to event bus
+    4. Orchestrators subscribe and coordinate downstream workflows
+
+    ```
+    Reducer -> Projection Commit -> Notification Publisher -> Event Bus
+                                            |
+                                            v
+                                    Orchestrators (subscribers)
+    ```
+
+Design Principles:
+    - **Loose Coupling**: Reducers don't know about orchestrators
+    - **At-Least-Once Delivery**: Consumers handle idempotency via projection_version
+    - **Circuit Breaker**: Resilience against event bus failures
+    - **Correlation Tracking**: Full distributed tracing support
+
+Concurrency Safety:
+    This implementation is coroutine-safe for concurrent async publishing.
+    Uses asyncio locks for circuit breaker state management. Note: This is
+    coroutine-safe, not thread-safe. For multi-threaded access, additional
+    synchronization would be required.
+
+Error Handling:
+    All methods raise ONEX error types:
+    - InfraConnectionError: Event bus unavailable or connection failed
+    - InfraTimeoutError: Publish operation timed out
+    - InfraUnavailableError: Circuit breaker open
+
+Example Usage:
+    ```python
+    from omnibase_infra.runtime import TransitionNotificationPublisher
+    from omnibase_core.models.notifications import ModelStateTransitionNotification
+
+    # Initialize publisher with event bus
+    publisher = TransitionNotificationPublisher(
+        event_bus=kafka_event_bus,
+        topic="onex.fsm.state.transitions.v1",
+    )
+
+    # Publish single notification
+    notification = ModelStateTransitionNotification(
+        aggregate_type="registration",
+        aggregate_id=entity_id,
+        from_state="pending",
+        to_state="active",
+        projection_version=1,
+        correlation_id=correlation_id,
+        causation_id=event_id,
+        timestamp=datetime.now(UTC),
+    )
+    await publisher.publish(notification)
+
+    # Batch publish
+    await publisher.publish_batch([notification1, notification2])
+
+    # Get metrics
+    metrics = publisher.get_metrics()
+    print(f"Published {metrics.notifications_published} notifications")
+    ```
+
+Related Tickets:
+    - OMN-1139: Implement TransitionNotificationPublisher
+
+See Also:
+    - ProtocolTransitionNotificationPublisher: Protocol definition (omnibase_core)
+    - ModelStateTransitionNotification: Notification model (omnibase_core)
+    - ProtocolEventBusLike: Event bus protocol
+"""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+import time
+from datetime import UTC, datetime
+from typing import TYPE_CHECKING, NamedTuple
+from uuid import UUID
+
+from omnibase_core.models.events.model_event_envelope import ModelEventEnvelope
+from omnibase_core.models.notifications import ModelStateTransitionNotification
+from omnibase_core.protocols.notifications import (
+    ProtocolTransitionNotificationPublisher,
+)
+from omnibase_core.utils.util_uuid_service import UtilUUID
+from omnibase_infra.enums import EnumInfraTransportType
+from omnibase_infra.errors import (
+    InfraConnectionError,
+    InfraTimeoutError,
+    InfraUnavailableError,
+    ModelInfraErrorContext,
+    ModelTimeoutErrorContext,
+)
+from omnibase_infra.mixins import MixinAsyncCircuitBreaker
+from omnibase_infra.models.resilience import ModelCircuitBreakerConfig
+from omnibase_infra.runtime.models.model_transition_notification_publisher_metrics import (
+    ModelTransitionNotificationPublisherMetrics,
+)
+from omnibase_infra.utils.util_error_sanitization import sanitize_error_string
+
+if TYPE_CHECKING:
+    from omnibase_infra.protocols import ProtocolEventBusLike
+
+logger = logging.getLogger(__name__)
+
+
+class FailedNotificationRecord(NamedTuple):
+    """Record of a failed notification publish attempt.
+
+    Used to track failures during batch publishing operations with clear
+    field semantics for error reporting and debugging.
+
+    Attributes:
+        aggregate_type: The type of aggregate that failed (e.g., "registration").
+        aggregate_id: The ID of the aggregate (as string for error reporting).
+        error_message: Sanitized error message describing the failure.
+    """
+
+    aggregate_type: str
+    aggregate_id: str
+    error_message: str
+
+
+class TransitionNotificationPublisher(MixinAsyncCircuitBreaker):
+    """Publishes transition notifications after projection commits.
+
+    Implements ProtocolTransitionNotificationPublisher from omnibase_core.
+    Provides at-least-once delivery semantics for state transition notifications
+    to enable orchestrator coordination without tight coupling to reducers.
+
+    Features:
+        - Protocol compliant (ProtocolTransitionNotificationPublisher)
+        - Circuit breaker resilience (MixinAsyncCircuitBreaker)
+        - Metrics tracking for observability
+        - Batch publishing for efficiency
+        - Correlation ID propagation for distributed tracing
+
+    Circuit Breaker:
+        Uses MixinAsyncCircuitBreaker for resilience:
+        - Opens after consecutive failures (configurable threshold)
+        - Resets after timeout period (configurable)
+        - Raises InfraUnavailableError when open
+
+    Thread Safety:
+        Coroutine-safe via asyncio.Lock for circuit breaker state.
+        Not thread-safe - use only from async context.
+
+    Attributes:
+        _event_bus: Event bus for publishing notifications
+        _topic: Target topic for notifications
+        _lock: Async lock for metrics updates
+        _publisher_id: Unique identifier for this publisher instance
+
+    Example:
+        >>> publisher = TransitionNotificationPublisher(event_bus, topic="notifications.v1")
+        >>> await publisher.publish(notification)
+        >>> metrics = publisher.get_metrics()
+        >>> print(f"Success rate: {metrics.publish_success_rate():.2%}")
+    """
+
+    # Default maximum number of failures to track in memory during batch operations.
+    # Prevents unbounded memory growth for very large batches with many failures.
+    # Can be overridden via constructor parameter for large batch tuning.
+    DEFAULT_MAX_TRACKED_FAILURES: int = 100
+
+    def __init__(
+        self,
+        event_bus: ProtocolEventBusLike,
+        topic: str,
+        *,
+        publisher_id: str | None = None,
+        circuit_breaker_threshold: int = 5,
+        circuit_breaker_reset_timeout: float = 60.0,
+        max_tracked_failures: int = DEFAULT_MAX_TRACKED_FAILURES,
+    ) -> None:
+        """Initialize transition notification publisher.
+
+        Args:
+            event_bus: Event bus implementing ProtocolEventBusLike for publishing.
+                Must support publish_envelope() method.
+            topic: Target topic for transition notifications. Required.
+                This should be configured in the projector's contract or
+                notification config rather than hardcoded. Example topics:
+                - "onex.fsm.state.transitions.v1"
+                - "registration.state.transitions.v1"
+            publisher_id: Optional unique identifier for this publisher instance.
+                If not provided, a UUID will be generated.
+            circuit_breaker_threshold: Maximum failures before opening circuit.
+                Default: 5
+            circuit_breaker_reset_timeout: Seconds before automatic reset.
+                Default: 60.0
+            max_tracked_failures: Maximum number of failures to track in memory
+                during batch operations. Prevents unbounded memory growth for
+                very large batches with many failures. For large batch operations,
+                this can be tuned higher to capture more failure details.
+                Default: 100
+
+        Example:
+            >>> publisher = TransitionNotificationPublisher(
+            ...     event_bus=kafka_event_bus,
+            ...     topic="onex.fsm.state.transitions.v1",
+            ...     circuit_breaker_threshold=3,
+            ...     circuit_breaker_reset_timeout=30.0,
+            ...     max_tracked_failures=200,  # Tune for large batches
+            ... )
+        """
+        self._event_bus = event_bus
+        self._topic = topic
+        self._publisher_id = (
+            publisher_id or f"transition-publisher-{UtilUUID.generate()!s}"
+        )
+        self._lock = asyncio.Lock()
+        self._max_tracked_failures = max_tracked_failures
+
+        # Metrics tracking
+        self._notifications_published = 0
+        self._notifications_failed = 0
+        self._batch_operations = 0
+        self._batch_notifications_attempted = 0
+        self._batch_notifications_total = 0
+        self._batch_failures_truncated = 0
+        self._last_publish_at: datetime | None = None
+        self._last_publish_duration_ms: float = 0.0
+        self._total_publish_duration_ms: float = 0.0
+        self._max_publish_duration_ms: float = 0.0
+        self._started_at = datetime.now(UTC)
+
+        # Initialize circuit breaker with configured settings
+        # Note: the mixin sets self.circuit_breaker_threshold and
+        # self.circuit_breaker_reset_timeout as instance attributes
+        cb_config = ModelCircuitBreakerConfig(
+            threshold=circuit_breaker_threshold,
+            reset_timeout_seconds=circuit_breaker_reset_timeout,
+            service_name=f"transition-notification-publisher.{topic}",
+            transport_type=EnumInfraTransportType.KAFKA,
+        )
+        self._init_circuit_breaker_from_config(cb_config)
+
+        logger.info(
+            "TransitionNotificationPublisher initialized",
+            extra={
+                "publisher_id": self._publisher_id,
+                "topic": self._topic,
+                "circuit_breaker_threshold": circuit_breaker_threshold,
+                "circuit_breaker_reset_timeout": circuit_breaker_reset_timeout,
+                "max_tracked_failures": self._max_tracked_failures,
+            },
+        )
+
+    @property
+    def topic(self) -> str:
+        """Get the configured topic."""
+        return self._topic
+
+    @property
+    def publisher_id(self) -> str:
+        """Get the publisher identifier."""
+        return self._publisher_id
+
+    async def publish(
+        self,
+        notification: ModelStateTransitionNotification,
+    ) -> None:
+        """Publish a single state transition notification.
+
+        Wraps the notification in a ModelEventEnvelope and publishes to the
+        configured topic via the event bus. Implements at-least-once delivery
+        semantics - consumers should handle idempotency via projection_version.
+
+        Args:
+            notification: The state transition notification to publish.
+
+        Raises:
+            InfraConnectionError: If event bus connection fails.
+            InfraTimeoutError: If publish operation times out.
+            InfraUnavailableError: If circuit breaker is open.
+
+        Example:
+            >>> notification = ModelStateTransitionNotification(
+            ...     aggregate_type="registration",
+            ...     aggregate_id=uuid4(),
+            ...     from_state="pending",
+            ...     to_state="active",
+            ...     projection_version=1,
+            ...     correlation_id=uuid4(),
+            ...     causation_id=uuid4(),
+            ...     timestamp=datetime.now(UTC),
+            ... )
+            >>> await publisher.publish(notification)
+        """
+        correlation_id = notification.correlation_id
+        start_time = time.monotonic()
+
+        # Check circuit breaker before operation
+        async with self._circuit_breaker_lock:
+            await self._check_circuit_breaker("publish", correlation_id)
+
+        ctx = ModelInfraErrorContext.with_correlation(
+            correlation_id=correlation_id,
+            transport_type=EnumInfraTransportType.KAFKA,
+            operation="publish_transition_notification",
+            target_name=self._topic,
+        )
+
+        try:
+            # Create envelope wrapping the notification model directly.
+            # ModelEventEnvelope[T] is generic and handles Pydantic models natively,
+            # serializing them lazily when needed via to_dict_lazy().
+            envelope = ModelEventEnvelope[ModelStateTransitionNotification](
+                payload=notification,
+                correlation_id=notification.correlation_id,
+                source_tool=self._publisher_id,
+            )
+
+            # Publish to event bus
+            await self._event_bus.publish_envelope(envelope, self._topic)
+
+            # Calculate duration
+            duration_ms = (time.monotonic() - start_time) * 1000
+
+            # Record success
+            async with self._circuit_breaker_lock:
+                await self._reset_circuit_breaker()
+
+            # Update metrics
+            async with self._lock:
+                self._notifications_published += 1
+                self._last_publish_at = datetime.now(UTC)
+                self._last_publish_duration_ms = duration_ms
+                self._total_publish_duration_ms += duration_ms
+                self._max_publish_duration_ms = max(
+                    self._max_publish_duration_ms, duration_ms
+                )
+
+            logger.debug(
+                "Published transition notification",
+                extra={
+                    "aggregate_type": notification.aggregate_type,
+                    "aggregate_id": str(notification.aggregate_id),
+                    "from_state": notification.from_state,
+                    "to_state": notification.to_state,
+                    "projection_version": notification.projection_version,
+                    "correlation_id": str(correlation_id),
+                    "duration_ms": duration_ms,
+                },
+            )
+
+        except (InfraUnavailableError, InfraTimeoutError):
+            # Re-raise infrastructure errors without wrapping - preserve error semantics
+            await self._handle_failure("publish", correlation_id)
+            raise
+
+        except TimeoutError as e:
+            await self._handle_failure("publish", correlation_id)
+            timeout_ctx = ModelTimeoutErrorContext(
+                transport_type=EnumInfraTransportType.KAFKA,
+                operation="publish_transition_notification",
+                target_name=self._topic,
+                correlation_id=correlation_id,
+            )
+            raise InfraTimeoutError(
+                f"Timeout publishing transition notification for "
+                f"{notification.aggregate_type}:{notification.aggregate_id}",
+                context=timeout_ctx,
+            ) from e
+
+        except Exception as e:
+            await self._handle_failure("publish", correlation_id)
+            raise InfraConnectionError(
+                f"Failed to publish transition notification for "
+                f"{notification.aggregate_type}:{notification.aggregate_id}",
+                context=ctx,
+            ) from e
+
+    async def publish_batch(
+        self,
+        notifications: list[ModelStateTransitionNotification],
+    ) -> None:
+        """Publish multiple state transition notifications.
+
+        Publishes each notification sequentially, continuing on individual
+        failures. This method is provided for efficiency when multiple
+        transitions occur in a single unit of work.
+
+        Ordering:
+            Notifications are published in the order provided. The order is
+            preserved when delivery order matters for workflow correctness.
+
+        Error Handling:
+            If any notification fails to publish, the error is raised after
+            attempting all notifications. Partial success is possible.
+
+        Circuit Breaker Behavior:
+            The circuit breaker is checked only at the start of the batch
+            operation. However, individual publish() calls within the batch
+            can trip the circuit breaker if they fail. If the circuit breaker
+            opens mid-batch (due to accumulated failures from individual
+            publish calls), subsequent notifications in the batch will fail
+            with InfraUnavailableError. This is expected "partial success"
+            behavior - the batch continues attempting all notifications, but
+            failures are recorded and reported at the end.
+
+        Correlation ID Behavior:
+            The batch uses the **first notification's correlation_id** for all
+            batch-level operations:
+
+            - Circuit breaker checks (at batch start)
+            - Batch summary logging ("Batch publish completed")
+            - Error context creation (when raising InfraConnectionError)
+            - Failure summary logging ("Batch publish failures - details")
+
+            However, **individual notification errors are logged with their own
+            correlation_id**. When a specific notification fails within the batch,
+            the warning log entry includes that notification's correlation_id,
+            not the batch correlation_id.
+
+            This design is intentional:
+
+            1. **Batch-level traceability**: Using a single correlation_id for
+               batch operations allows operators to correlate all batch-related
+               log entries and metrics under one trace ID.
+
+            2. **Per-notification traceability**: Individual failure logs retain
+               their specific correlation_id, enabling operators to trace the
+               complete lifecycle of each notification independently.
+
+            Example log correlation::
+
+                # Batch-level log (uses first notification's correlation_id)
+                {"message": "Batch publish completed", "correlation_id": "aaa-111"}
+
+                # Individual failure log (uses that notification's correlation_id)
+                {"message": "Failed to publish notification in batch",
+                 "correlation_id": "bbb-222"}
+
+        Args:
+            notifications: List of notifications to publish.
+
+        Raises:
+            InfraConnectionError: If event bus connection fails.
+            InfraTimeoutError: If publish operation times out.
+            InfraUnavailableError: If circuit breaker is open (at batch start
+                or if tripped mid-batch by individual publish failures).
+
+        Example:
+            >>> notifications = [notification1, notification2, notification3]
+            >>> await publisher.publish_batch(notifications)
+        """
+        if not notifications:
+            return
+
+        correlation_id = notifications[0].correlation_id
+        start_time = time.monotonic()
+
+        # Batch-level circuit breaker check for fail-fast behavior.
+        # NOTE: This check is NOT redundant with the per-notification check in publish().
+        # - This check: Fail-fast before starting any work if circuit is already open
+        # - Per-notification checks in publish(): Handle circuit opening MID-batch due to
+        #   accumulated failures during batch processing (expected partial-success behavior)
+        # See docstring "Circuit Breaker Behavior" section for full explanation.
+        async with self._circuit_breaker_lock:
+            await self._check_circuit_breaker("publish_batch", correlation_id)
+
+        success_count = 0
+        last_error: Exception | None = None
+        failed_notifications: list[FailedNotificationRecord] = []
+        truncation_occurred = False
+        # Track error types to determine most severe error for final raise.
+        # Severity order: InfraUnavailableError > InfraTimeoutError > InfraConnectionError
+        encountered_unavailable = False
+        encountered_timeout = False
+
+        for notification in notifications:
+            try:
+                await self.publish(notification)
+                success_count += 1
+            except (
+                InfraConnectionError,
+                InfraTimeoutError,
+                InfraUnavailableError,
+            ) as e:
+                last_error = e
+                # Track error types for determining most severe error to raise
+                if isinstance(e, InfraUnavailableError):
+                    encountered_unavailable = True
+                elif isinstance(e, InfraTimeoutError):
+                    encountered_timeout = True
+                # Only track failures up to the limit to prevent unbounded memory growth
+                if len(failed_notifications) < self._max_tracked_failures:
+                    failed_notifications.append(
+                        FailedNotificationRecord(
+                            aggregate_type=notification.aggregate_type,
+                            aggregate_id=str(notification.aggregate_id),
+                            error_message=sanitize_error_string(str(e)),
+                        )
+                    )
+                else:
+                    # Mark that truncation occurred (limit reached)
+                    truncation_occurred = True
+                logger.warning(
+                    "Failed to publish notification in batch",
+                    extra={
+                        "aggregate_type": notification.aggregate_type,
+                        "aggregate_id": str(notification.aggregate_id),
+                        "error": sanitize_error_string(str(e)),
+                        "correlation_id": str(notification.correlation_id),
+                    },
+                )
+                # Continue with remaining notifications
+
+        # Calculate duration
+        duration_ms = (time.monotonic() - start_time) * 1000
+
+        # Update batch metrics
+        async with self._lock:
+            self._batch_operations += 1
+            self._batch_notifications_attempted += len(notifications)
+            self._batch_notifications_total += success_count
+            if truncation_occurred:
+                self._batch_failures_truncated += 1
+
+        failure_count = len(notifications) - success_count
+
+        # Log aggregate failure information when truncation occurs
+        if truncation_occurred:
+            failure_summary = self._summarize_failure_types(failed_notifications)
+            untracked_failures = failure_count - len(failed_notifications)
+            logger.warning(
+                "Batch publish failure tracking truncated",
+                extra={
+                    "correlation_id": str(correlation_id),
+                    "total_failures": failure_count,
+                    "tracked_failures": len(failed_notifications),
+                    "untracked_failures": untracked_failures,
+                    "max_tracked_failures": self._max_tracked_failures,
+                    "failure_type_summary": failure_summary,
+                },
+            )
+
+        logger.info(
+            "Batch publish completed",
+            extra={
+                "total": len(notifications),
+                "success": success_count,
+                "failed": failure_count,
+                "duration_ms": duration_ms,
+                "correlation_id": str(correlation_id),
+            },
+        )
+
+        # Raise with detailed failure information if any failures occurred
+        if last_error is not None:
+            ctx = ModelInfraErrorContext.with_correlation(
+                correlation_id=correlation_id,
+                transport_type=EnumInfraTransportType.KAFKA,
+                operation="publish_batch",
+                target_name=self._topic,
+            )
+
+            # Log failure details for debugging before raising truncated error.
+            # Limit logged failures to prevent oversized log entries while
+            # preserving full counts for metrics and observability.
+            max_logged_failures = 10
+            logged_failures = [
+                {
+                    "aggregate_type": record.aggregate_type,
+                    "aggregate_id": record.aggregate_id,
+                    "error_message": record.error_message,
+                }
+                for record in failed_notifications[:max_logged_failures]
+            ]
+            failures_truncated = len(failed_notifications) > max_logged_failures
+
+            logger.warning(
+                "Batch publish failures - details",
+                extra={
+                    "correlation_id": str(correlation_id),
+                    "topic": self._topic,
+                    "total_notifications": len(notifications),
+                    "success_count": success_count,
+                    "failure_count": failure_count,
+                    "tracked_failures": len(failed_notifications),
+                    "max_tracked_failures": self._max_tracked_failures,
+                    "logged_failures": len(logged_failures),
+                    "failures_truncated": failures_truncated,
+                    "failures": logged_failures,
+                },
+            )
+
+            # Build detailed error message showing first 3 failures
+            failure_details = "; ".join(
+                f"{record.aggregate_type}:{record.aggregate_id[:8]}... - "
+                f"{record.error_message[:50]}"
+                for record in failed_notifications[:3]
+            )
+            if failure_count > 3:
+                failure_details += f" ... and {failure_count - 3} more"
+
+            error_message = (
+                f"Batch publish partially failed: {failure_count}/{len(notifications)} "
+                f"notifications failed ({success_count} succeeded). "
+                f"Failures: [{failure_details}]"
+            )
+
+            # Raise the most severe error type encountered during batch processing.
+            # Severity order: InfraUnavailableError > InfraTimeoutError > InfraConnectionError
+            # This preserves error semantics so callers can handle appropriately
+            # (e.g., retry on timeout, skip on unavailable).
+            if encountered_unavailable:
+                raise InfraUnavailableError(
+                    error_message,
+                    context=ctx,
+                ) from last_error
+            if encountered_timeout:
+                timeout_ctx = ModelTimeoutErrorContext(
+                    transport_type=EnumInfraTransportType.KAFKA,
+                    operation="publish_batch",
+                    target_name=self._topic,
+                    correlation_id=correlation_id,
+                )
+                raise InfraTimeoutError(
+                    error_message,
+                    context=timeout_ctx,
+                ) from last_error
+            raise InfraConnectionError(
+                error_message,
+                context=ctx,
+            ) from last_error
+
+    async def _handle_failure(
+        self,
+        operation: str,
+        correlation_id: UUID,
+    ) -> None:
+        """Handle a publish failure by recording circuit breaker failure.
+
+        Args:
+            operation: Operation name for logging
+            correlation_id: Correlation ID for tracing
+        """
+        async with self._circuit_breaker_lock:
+            await self._record_circuit_failure(operation, correlation_id)
+
+        async with self._lock:
+            self._notifications_failed += 1
+
+    def _summarize_failure_types(
+        self, failures: list[FailedNotificationRecord]
+    ) -> dict[str, int]:
+        """Summarize failure types by grouping error messages.
+
+        Groups failures by a simplified error pattern (first 50 characters of
+        the error message) to help operators understand what types of errors
+        are occurring, even when detailed failure records are truncated.
+
+        Args:
+            failures: List of failed notification records to summarize.
+
+        Returns:
+            Dictionary mapping error pattern (truncated error message) to
+            the count of failures with that pattern.
+
+        Example:
+            >>> failures = [
+            ...     FailedNotificationRecord("reg", "id1", "Connection refused to broker"),
+            ...     FailedNotificationRecord("reg", "id2", "Connection refused to broker"),
+            ...     FailedNotificationRecord("reg", "id3", "Timeout waiting for response"),
+            ... ]
+            >>> summary = publisher._summarize_failure_types(failures)
+            >>> # {"Connection refused to broker": 2, "Timeout waiting for response": 1}
+        """
+        summary: dict[str, int] = {}
+        for failure in failures:
+            # Use first 50 chars as the pattern key for grouping
+            pattern = failure.error_message[:50]
+            summary[pattern] = summary.get(pattern, 0) + 1
+        return summary
+
+    def get_metrics(self) -> ModelTransitionNotificationPublisherMetrics:
+        """Get current publisher metrics.
+
+        Returns a snapshot of the publisher's operational metrics including
+        notification counts, timing information, and circuit breaker state.
+
+        Returns:
+            ModelTransitionNotificationPublisherMetrics with current values.
+
+        Example:
+            >>> metrics = publisher.get_metrics()
+            >>> print(f"Published: {metrics.notifications_published}")
+            >>> print(f"Success rate: {metrics.publish_success_rate():.2%}")
+            >>> print(f"Healthy: {metrics.is_healthy()}")
+        """
+        # Get circuit breaker state
+        cb_state = self._get_circuit_breaker_state()
+        cb_open = cb_state.get("state") == "open"
+        failures_value = cb_state.get("failures", 0)
+        consecutive_failures = failures_value if isinstance(failures_value, int) else 0
+
+        # Calculate average duration (only from successful publishes since
+        # _total_publish_duration_ms is only updated on success)
+        average_duration = (
+            self._total_publish_duration_ms / self._notifications_published
+            if self._notifications_published > 0
+            else 0.0
+        )
+
+        return ModelTransitionNotificationPublisherMetrics(
+            publisher_id=self._publisher_id,
+            topic=self._topic,
+            notifications_published=self._notifications_published,
+            notifications_failed=self._notifications_failed,
+            batch_operations=self._batch_operations,
+            batch_notifications_attempted=self._batch_notifications_attempted,
+            batch_notifications_total=self._batch_notifications_total,
+            batch_failures_truncated=self._batch_failures_truncated,
+            last_publish_at=self._last_publish_at,
+            last_publish_duration_ms=self._last_publish_duration_ms,
+            average_publish_duration_ms=average_duration,
+            max_publish_duration_ms=self._max_publish_duration_ms,
+            circuit_breaker_open=cb_open,
+            consecutive_failures=consecutive_failures,
+            started_at=self._started_at,
+        )
+
+
+# Protocol compliance check (runtime_checkable allows isinstance checks)
+def _verify_protocol_compliance() -> None:  # pragma: no cover
+    """Verify TransitionNotificationPublisher implements the protocol.
+
+    This function is never called at runtime - it exists purely for static
+    type checking verification that the implementation is protocol-compliant.
+    """
+    from typing import cast
+
+    from omnibase_infra.event_bus.event_bus_inmemory import EventBusInmemory
+
+    # Create instance to verify protocol compliance
+    bus = cast("ProtocolEventBusLike", EventBusInmemory())
+    publisher: ProtocolTransitionNotificationPublisher = (
+        TransitionNotificationPublisher(
+            event_bus=bus,
+            topic="onex.fsm.state.transitions.v1",
+        )
+    )
+    # Use the variable to silence unused warnings
+    _ = publisher
+
+
+__all__: list[str] = ["FailedNotificationRecord", "TransitionNotificationPublisher"]