- Add BroadcasterMetrics interface for reading metrics per namespace
- Add MetricsCollector interface and DefaultMetricsCollector implementation
- Track events_published and events_received counters per namespace
- Track active_subscriptions gauge per namespace
- Track publish_errors, subscribe_errors, and dropped_events counters
- Add MetricsProvider interface for EventBroadcaster implementations
- Integrate metrics tracking into EventBus and NATSEventBus
- Add optional Prometheus integration via PrometheusMetricsAdapter
- Add comprehensive unit tests for metrics functionality
Closes#22
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add comprehensive integration tests that verify NATSEventBus behavior
with a real NATS server. Tests cover:
- Cross-node event delivery (multiple NATSEventBus instances)
- Namespace isolation with single and multiple NATS connections
- High-throughput scenarios (1000 events)
- Event ordering within namespace
- No cross-namespace leakage verification
- Concurrent publish/subscribe operations
- Multiple subscribers to same namespace
- Event metadata preservation across NATS
- Large event payload handling (100KB)
- Subscribe/unsubscribe lifecycle
- Reconnection behavior
- Graceful degradation under load
- Benchmarks for publish and publish-receive
Tests require a running NATS server and are tagged with +build integration.
Run with: go test -tags=integration -v ./...
Closes#18
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Support NATS-style wildcard patterns ("*" and ">") for subscribing
to events across multiple namespaces. This enables cross-cutting
concerns like logging, monitoring, and auditing without requiring
separate subscriptions for each namespace.
- Add pattern.go with MatchNamespacePattern and IsWildcardPattern
- Update EventBus to track wildcard subscribers separately
- Update NATSEventBus to use NATS native wildcard support
- Add comprehensive tests for pattern matching and EventBus wildcards
- Document security implications in all relevant code comments
Closes#20
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add a README.md that gives developers a quick understanding of what
Aether is and how to get started. Includes:
- Project description and why Aether exists
- Installation instructions
- Quick start code example showing event creation, persistence, and replay
- Key concepts (immutability, derived state, version consistency)
- Links to further documentation
- CI badge
Closes#44
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add support for optional namespace prefixes on JetStreamEventStore streams
to enable complete namespace isolation at the storage level:
- Add Namespace field to JetStreamConfig
- Add NewJetStreamEventStoreWithNamespace convenience constructor
- Prefix stream names with sanitized namespace when configured
- Add GetNamespace and GetStreamName accessor methods
- Add unit tests for namespace functionality
- Document namespace-scoped stores in CLAUDE.md
The namespace prefix is sanitized (spaces, dots, wildcards converted to
underscores) and prepended to the stream name, ensuring events from one
namespace cannot be read from another namespace's store while maintaining
full backward compatibility for non-namespaced stores.
Closes#19
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add ReplayError and ReplayResult types to capture information about
malformed events encountered during replay. This allows callers to
inspect and handle corrupted data rather than having it silently skipped.
Key changes:
- Add ReplayError type with sequence number, raw data, and underlying error
- Add ReplayResult type containing both successfully parsed events and errors
- Add EventStoreWithErrors interface for stores that can report replay errors
- Implement GetEventsWithErrors on JetStreamEventStore
- Update GetEvents to maintain backward compatibility (still skips malformed)
- Add comprehensive unit tests for the new types
This addresses the issue of silent data loss during event-sourced replay
by giving callers visibility into data quality issues.
Closes#39
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add sync.RWMutex to ConsistentHashRing struct
- Use Lock/Unlock for write operations (AddNode, RemoveNode)
- Use RLock/RUnlock for read operations (GetNode, GetNodes, IsEmpty)
This allows concurrent reads (the common case) while serializing writes,
preventing race conditions when multiple goroutines access the hash ring.
Closes#35
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
JSON unmarshal produces map[string]interface{}, not concrete types.
Added ModelPayload and MessagePayload concrete types that implement
RuntimeModel and RuntimeMessage interfaces respectively.
The handleClusterMessage now re-marshals and unmarshals the payload
to convert from map[string]interface{} to the proper concrete type.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add VirtualMachine interface with GetID(), GetActorID(), GetState()
- Add VMState type with idle/running/paused/stopped states
- Add RuntimeModel interface for event storming model contracts
- Add RuntimeMessage interface for actor message contracts
- Add VMProvider interface for decoupled VM access
- Update VMRegistry.GetActiveVMs() to return map[string]VirtualMachine
- Update Runtime interface to use RuntimeModel and RuntimeMessage
- Update DistributedVMRegistry to use VMProvider instead of interface{}
- Add SetVMProvider method to DistributedVM for dependency injection
This improves type safety, makes contracts explicit, and enables better
IDE support while avoiding import cycles through interface segregation.
Closes#37
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add config structs with sensible defaults for tunable parameters:
- JetStreamConfig for stream retention (1 year) and replica count (1)
- HashRingConfig for virtual nodes per physical node (150)
- ShardConfig for shard count (1024) and replication factor (1)
Each component gets a new WithConfig constructor that accepts custom
configuration, while the original constructors continue to work with
defaults. Zero values in configs fall back to defaults for backward
compatibility.
Closes#38
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add Metadata field (map[string]string) to Event struct with omitempty
- Add helper methods for common metadata: SetCorrelationID/GetCorrelationID,
SetCausationID/GetCausationID, SetUserID/GetUserID, SetTraceID/GetTraceID,
SetSpanID/GetSpanID
- Add WithMetadataFrom helper for copying metadata between events
- Add metadata key constants for standard fields
- Add comprehensive unit tests for metadata serialization and helpers
- Add store tests verifying metadata persistence
Closes#7
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add ErrVersionConflict error type and VersionConflictError for detailed
conflict information
- Implement version validation in InMemoryEventStore.SaveEvent that rejects
events with version <= current latest version
- Implement version validation in JetStreamEventStore.SaveEvent with version
caching for performance
- Add comprehensive tests for version conflict detection including concurrent
writes to same actor
- Document versioning semantics in EventStore interface and CLAUDE.md
This ensures events have monotonically increasing versions per actor and
provides clear error messages for version conflicts, enabling optimistic
concurrency control patterns.
Closes#6
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Test SaveEvent persists events correctly (single, multiple, multi-actor)
- Test GetEvents retrieves events in insertion order
- Test GetEvents with fromVersion filtering
- Test GetLatestVersion returns correct version
- Test behavior with non-existent actor IDs (returns empty/zero)
- Test concurrent access safety with race detector
- Add mutex protection to InMemoryEventStore for thread safety
Closes#3
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Test all public methods with 100% coverage:
- AddNode: updates ring, is idempotent, handles multiple nodes
- RemoveNode: updates ring, handles non-existent nodes
- GetNode: returns consistent results, handles empty ring and single node
- GetNodes and IsEmpty helper methods
Distribution tests verify:
- Balanced key distribution across nodes (< 25% deviation)
- Minimal key movement when adding nodes (< 35% moved)
- Virtual nodes improve distribution (CV < 15%)
- Ring behavior with 100+ nodes
Includes benchmarks for GetNode, AddNode, and distribution operations.
Closes#2
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Test JSON serialization/deserialization, field names, omitempty behavior,
edge cases (empty/nil data, large payloads, special characters including
unicode and control chars), timestamp handling across timezones, nanosecond
precision, version edge cases, and nested data structures.
Closes#1
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
