842cf7c162
* Avoid null endpoint DTO metadata in tests * Enforce console logs hub read permission * Remove unused console logs hub import * Support mapped endpoint metadata in auth tests * Reduce console log capture throughput impact * Address Copilot console logs review * Refactor task scheduling to support tenant-level background work and enhance logging functionality. * Introduce ConsoleStreamHook for stdout/stderr tee and enhance logging validation. Adjust test cases and startup warnings for distributed lock provider usage. * Refactor console logging pipeline with capture optimization and new ConsoleLogsHost; update tests accordingly. * Add Ansi SGR parser for console logs and associated unit tests * Remove ANSI color renderings and parsers; integrate ConsoleLogScopeAccessor for improved logging context with workflow instance ID support. * Address console logs code quality feedback * Address PR review feedback * Preserve console logs extension points * Stabilize console logs host lifecycle * Address final automated review comments * Tighten console log capture shutdown * Address console log review feedback * Address follow-up review feedback * Cover final review feedback * Avoid recursive console provider initialization * Guard console host lease shutdown * Preserve console log scope and provider lifetime * Correlate console log scope fallback * Tighten console scope correlation * Expose host services during provider construction * Redact ANSI-normalized console lines * Add OpenTelemetry diagnostics backend foundation * Add OTLP HTTP ingestion parsing * Document OpenTelemetry diagnostics setup * Enforce OpenTelemetry hub permissions * Remove `ConsoleCaptureTee` and related services and tests * Add OpenTelemetry HTTP ingestion integration test * Use pipeline contributors for console log context * Update CShells package versions to 0.0.24-preview.132 * Add OpenTelemetry ingestion security tests * Add OpenTelemetry API authorization tests * Filter live console logs by workflow instance * Add OpenTelemetry hub tests * Add OpenTelemetry gRPC metadata hook * Assert OpenTelemetry workflow tags survive ingestion * Mark OpenTelemetry core build verified * Enhance console logging with activity execution metadata and extend test coverage. * Address console logs stream consumption comment * Wire OpenTelemetry diagnostics into core sample * Address Core diagnostics review feedback * Address Core Copilot follow-up feedback * Add OpenTelemetry metric instrument names * Address Core Copilot provider feedback * Address Core Copilot diagnostics follow-up * Address Core Copilot live feed feedback * Address Core Copilot store feedback * Integrate OpenTelemetry for logging, tracing, and metrics in ModularServer and update launch settings and docker-compose configuration. * Refactor to replace `ConsoleLogStream.Core` with `ConsoleLogStreaming.Core` across codebase and update `ConsoleStreamHook` installation. * Add diagnostics OpenTelemetry backend * Fix OpenTelemetry live hub subscription * Fix modular OpenTelemetry exporter endpoints * Add CShells logging configuration in appsettings.json * Remove obsolete unit tests and helper classes * Restore default activity exception handling * Simplify type serialization and alias management This commit refactors the internal type serialization and alias management system to reduce boilerplate, improve robustness, and simplify the developer experience: - Removed numerous explicit `ExpressionOptions` type alias registrations across various modules. - Updated `TypeJsonConverter` and polymorphic serialization to reliably handle types using assembly-qualified names when a short alias is not explicitly registered. - Streamlined `ExcludeFromHashConverter` to strictly adhere to `ExcludeFromHashAttribute` for hash calculations, removing complex `JsonIgnoreCondition` logic. - Eliminated several helper classes (`WorkflowJsonTypeResolver`, `WorkflowTypeValidator`, `IWorkflowTypeRegistry`, `WorkflowFactoryDictionary`, `JavaScriptExceptionTypeAliasRegistrar`, `WorkflowRuntimeTypeAliasRegistrar`) and their associated unit tests, simplifying the codebase. Additionally, this commit introduces a comprehensive markdown document (`product-website-feature-source.md`) outlining Elsa's core features, Studio capabilities, extension ecosystem, and architectural selling points, intended as source material for the product website. * Refine type serialization for improved robustness and alias handling This commit further enhances the type serialization and deserialization mechanisms: * Centralizes type resolution and alias management through `IWellKnownTypeRegistry` and `WorkflowJsonTypeResolver`. * Prioritizes registered type aliases when serializing type metadata in `PolymorphicObjectConverter`, resulting in more concise JSON output. * Enhances deserialization in `PolymorphicObjectConverter` and `VariableMapper` to gracefully handle unknown or non-instantiable types, providing fallbacks and logging warnings. * Simplifies `TypeJsonConverter` by delegating complex type resolution logic to the `WorkflowJsonTypeResolver`. * Adds `JsonArray` to the well-known type aliases for direct recognition. * Fix console logs packaging and workflow type resolution * Fix console log metadata and type resolution * Address Copilot review feedback * Enhance type resolution, improve console log handling, and update tests - Streamlined `WorkflowDictionaryExtensions` for better workflow registration validation. - Refined `ConsoleLogsAuthorizationTests` with the new `SetJsonRequest` helper to improve test requests handling. - Updated `OrderDefinition` to ignore JSON serialization for `KeySelector`. - Enhanced `WorkflowRuntimeFeature` for improved workflow registration and type alias configuration. - Added tests to ensure `ConsoleLogProvider` metadata filtration in various scenarios. - Improved type serialization logic in `WorkflowJsonTypeResolver`. - Updated README to fix references related to diagnostics. - Optimized `ExcludeFromHashConverter` for property serialization conditions. - Modified `TriggerIndexer` for streamlined trigger management. - Tested payload checks in `PublishEventTests`. - Adjusted `Endpoint` in `ConsoleLogs` for automatic JSON request handling. - Ensured registration of workflow type aliases in `WorkflowsFeature`. * Restore CLR workflow registration compatibility * Align JSON island serialization fixtures * Add Console Logs Services and Enhance Endpoint Handling - Introduced `ActivityExecutionsEndpointTests` to validate route exposure. - Added `ConsoleLogCaptureHostedService` for console log streaming. - Implemented `ConsoleStreamJsonConverter` for JSON conversion of console streams. - Developed `ElsaConsoleLogRecentBuffer` to handle recent log buffering. - Updated `ConsoleLogsAuthorizationTests` with new test cases for stream filter mapping. - Consolidated console log provider dependencies and registration, including recent buffering. - Enhanced `ElsaConsoleLogProvider` to use recent buffer for filtering. - Adjusted `Program.cs` for streamlined logging service setup. * Enhance type resolution and test coverage; streamline console log integration - Added `ConsoleStreamHook` for streamlined log streaming. - Updated `WorkflowJsonTypeResolverTests` to improve type resolution and test new scenarios. - Simplified type resolution by removing trusted assembly checks. * Fix CI smoke and package restore failures * Fix Docker smoke image project paths * Fix Docker Python runtime packages * Fix Docker CA smoke teardown * Refresh Elsa roadmap * Implement background processors and mediation coordination - Added `BackgroundCommandProcessor`, `BackgroundJobProcessor`, and `BackgroundNotificationProcessor` classes for handling commands, jobs, and notifications, respectively. - Introduced `MediatorBackgroundProcessingCoordinator` to coordinate the execution of all background processors. - Implemented `MediatorBackgroundTask` for wrapping `MediatorBackgroundProcessingCoordinator` in `BackgroundTask`. - Added unit tests for `MediatorBackgroundTask` to ensure proper start and stop behavior. - Refactored `BackgroundCommandSenderHostedService` to utilize `BackgroundCommandProcessor`. - Introduced 'elsa-roadmap-refresh' skill configuration for roadmap updates. * Address workflow type resolution review feedback * Address follow-up review feedback * Restore recent console logs execute path * Address Copilot follow-up review * Decouple workflow JSON aliases from expressions * Fix workflow management unit test setup * Fix console logs recent endpoint handler shape * Respect workflow JSON strict type aliases * Remove unused console log contracts reference * Address Copilot review feedback * Address Copilot follow-up comments * Synchronize ring buffer dropped count * Address background processor strategy replay
3 KiB
3 KiB
Research: Diagnostics OpenTelemetry
Decision: Core owns the collector/API boundary
Rationale: OTLP ingestion must happen in an ASP.NET Core backend, not in Studio. Core can authenticate diagnostics APIs, protect ingestion, normalize protobuf payloads, and stream live updates.
Alternatives considered:
- Studio collector: rejected because browser/Studio hosts cannot reliably expose OTLP endpoints.
- External collector only: rejected because this feature needs a first-party local diagnostics story.
Decision: Preserve existing workflow instrumentation
Rationale: Current Elsa.Workflows.Core already emits Elsa.Workflows spans and metrics through System.Diagnostics. This module should consume/export that telemetry through OTLP rather than re-instrument workflow execution.
Alternatives considered:
- Port
elsa-extensionsElsa.OpenTelemetry: rejected because it is producer-side middleware, overlaps current Core instrumentation, and includes trace-boundary behavior that mutates ambientActivity.Current. - Move instrumentation into diagnostics: rejected because workflow instrumentation is valuable even when diagnostics collector is not installed.
Decision: HTTP/protobuf required, gRPC optional
Rationale: HTTP/protobuf fits existing ASP.NET Core endpoint patterns and common deployment routes. gRPC is useful for SDK defaults but depends on host configuration and package choices, so it should be enabled when available and represented honestly in collector metadata.
Alternatives considered:
- gRPC only: rejected because it complicates proxying and local HTTP testing.
- Always return a gRPC endpoint: rejected because it misleads clients when the host does not expose gRPC.
Decision: Bounded in-memory v1 storage
Rationale: Traces, metrics, and logs can be high volume. A bounded in-memory default matches the existing diagnostics modules and keeps v1 focused.
Alternatives considered:
- Durable storage now: rejected because trace and metric schemas need a deliberate follow-up provider design.
- Raw OTLP payload storage only: rejected because diagnostics clients need filtered, assembled trace and metric models.
Decision: OTLP logs stay separate from Structured Logs
Rationale: OTLP logs can come from arbitrary resources and use OTEL schemas. Elsa.Diagnostics.StructuredLogs owns first-party ILogger capture. Correlation should use trace/span IDs, not shared storage.
Alternatives considered:
- Merge OTLP logs into Structured Logs: rejected because schemas, retention, permissions, and source semantics differ.
Decision: Explicit non-loopback protection
Rationale: OTLP ingestion can expose sensitive application metadata and create high traffic. Loopback development can be low friction, but remote ingestion needs an API key header or equivalent configured protection.
Alternatives considered:
- No OTLP sender auth: rejected as unsafe.
- Require interactive user auth for OTLP senders: rejected because SDKs usually support headers, not browser auth.