Draft
Conversation
Gap analysis comparing C# SDK against the Python SDK reference implementation, covering bugs, missing clients, task types, worker framework, metrics, events, testing, examples, and documentation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1. Fix namespace typo in LlmIndexText.cs (DefinitaskNametion -> Definition) 2. Fix duplicate key bug in LlmChatComplete.cs (MAXTOKENS -> STOPWORDS for StopWords) 3. Fix inverted cancellation check in WorkflowTaskExecutor.cs (== -> !=) 4. Preserve stack trace in WorkflowTaskService.cs (throw ex -> throw) 5. Fix filename typo IWorkflowTaskCoodinator -> IWorkflowTaskCoordinator 6. Fix broken using in VectorDbHelloWorld.cs (cascading from #1) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Interfaces: IWorkflowClient, ITaskClient, IMetadataClient, ISchedulerClient, ISecretClient, IAuthorizationClient, IPromptClient, IIntegrationClient, IEventClient - each wrapping the raw API clients with cleaner APIs. Orkes implementations delegate to the underlying auto-generated API clients. OrkesClients factory creates all clients from a single Configuration. Void-returning async API methods are wrapped with Task.Run() to provide proper Task-based async signatures on the high-level interfaces. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ow, Inline, LlmStoreEmbeddings, LlmSearchEmbeddings, GetDocument, GenerateImage, GenerateAudio, ListMcpTools, CallMcpTool, ToolCall, ToolSpec Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…, lease extension, pause/resume - Adaptive exponential backoff on empty poll queues (configurable multiplier and max interval) - Worker health check tracking (consecutive errors, task counts, timestamps) - IsHealthy() and GetHealthStatuses() on coordinator for monitoring - Lease extension timer for long-running tasks (> configurable threshold) - Runtime pause/resume via environment variable checking - New configuration options in WorkflowTaskExecutorConfiguration Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…rics Provides counters and histograms for: - Task polling (count, success, empty, error, latency) - Task execution (count, success, error, latency) - Task updates (count, error, retry, latency) - Worker restarts and payload sizes - API calls (count, error, latency) Includes TimingScope helper for latency measurement. Integrates with OpenTelemetry and Prometheus exporters. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- ITaskRunnerEventListener: poll, execution, and update lifecycle events - IWorkflowEventListener: workflow start, complete, fail, terminate, pause, resume events - EventDispatcher: singleton with thread-safe listener registration and dispatch - Safe exception handling in dispatch to prevent listener errors from affecting workers Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
LLM Providers: Anthropic, AWS Bedrock, Cohere, Grok, Mistral, Ollama, Perplexity Vector DBs: PostgreSQL/pgvector, MongoDB Each provider has enum value and configuration class with environment variable support. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Test coverage for: - Phase 1: Bug fix regression tests (namespace, stopwords key, cancellation, stacktrace, interface name) - Phase 2: OrkesClients factory and client interface tests - Phase 3: All 13 new task type DSL tests (HttpPoll, Kafka, StartWorkflow, Inline, LLM tasks, MCP, Tool) - Phase 4: Worker monitor health checks, executor configuration defaults - Phase 5: Metrics counters, histograms, timing scope, MeterListener integration - Phase 6: Event dispatcher registration, unregistration, multi-listener, error isolation - Phase 7: AI provider enums, config classes for all 9 new providers + 2 vector DBs All 106 tests pass. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…async dispatch Phase 4: Auto-restart with configurable max retries and delay, 3-tier configuration (code < global env vars < worker-specific env vars) Phase 5: MetricsConfig for toggling metric categories, WorkerMetrics per-worker helper with consistent tagging and timing scopes Phase 6: Async event dispatch methods (DispatchTaskRunnerEventAsync, DispatchWorkflowEventAsync) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Examples added: - KitchenSink: all task types in one workflow - WorkerConfiguration: worker config with annotations, env overrides - EventListener: event listener registration and dispatch - Metrics: metrics recording, MeterListener, TimingScope - HumanInLoopChat: human-in-the-loop chat with LLM + HumanTask - MultiAgentChat: multi-agent collaboration (research, writing, review) - WorkflowTestExample: workflow testing with TaskMock - WorkerDiscovery: assembly scanning for IWorkflowTask implementations - AspNetCoreIntegration: DI patterns, controllers, BackgroundService - MetadataJourney: full metadata CRUD lifecycle - ScheduleJourney: scheduler operations - PromptJourney: prompt template CRUD and testing - AuthorizationJourney: apps, users, groups, permissions - McpAgentExample: MCP agent workflow - RagPipelineExample: RAG ingestion and query pipelines - WorkflowOps: complete workflow lifecycle operations Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New test files: - SerializationTests: 30 tests for model JSON round-trip (WorkflowDef, TaskDef, WorkflowTask, Task, TaskResult, StartWorkflowRequest, Workflow, WorkflowStatus, SubWorkflowParams, TaskMock, WorkflowTestRequest) - HighLevelClientTests: 25 tests for mocked client interfaces (IWorkflowClient, ITaskClient, IMetadataClient, ISchedulerClient, ISecretClient, IPromptClient, IAuthorizationClient) - WorkerFrameworkTests: 20 tests for worker config edge cases (exponential backoff, auto-restart, lease extension, pause/resume, 3-tier config, health status) - WorkflowBuilderTests: 31 tests for ConductorWorkflow builder and all task type DSLs (Simple, HTTP, HttpPoll, Switch, ForkJoin, DoWhile/Loop, SubWorkflow, Wait, Terminate, Inline, JQ, Kafka, StartWorkflow, LLM tasks, MCP tasks) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- README.md: comprehensive rewrite with quick-start, features overview, task types table, AI/LLM section, examples table, documentation links - workers.md: expanded with all worker features (auto-discovery, annotated workers, health checks, event listeners, metrics) - workflow.md: expanded with all task types, workflow operations, workflow testing, AI/LLM workflows, RAG pipeline, MCP agent - worker_configuration.md: new guide covering exponential backoff, auto-restart, lease extension, pause/resume, 3-tier config, health checks - metrics.md: new guide covering all metrics, WorkerMetrics helper, MetricsConfig, MeterListener, OpenTelemetry/Prometheus integration Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…uto-registration
- task-update-v2: UpdateTaskV2 on TaskResourceApi (PUT /tasks/{taskId}) returns next task
in one call; WorkflowTaskHttpClient falls back to v1 on 404/405/501 with a sticky flag
so the server is only probed once; WorkflowTaskExecutor feeds the returned next task
directly into ProcessTask, skipping an extra poll round-trip
- OSS auth auto-disable: TokenHandler sets _authDisabled on 404/405 from the token
endpoint and returns null for all subsequent requests, enabling zero-config OSS usage
- Interceptors: new IConductorInterceptor interface (BeforeRequest/AfterResponse);
Configuration.Interceptors list wired into ApiClient sync and async call paths
- Auto schema registration: [WorkerTask] gains RegisterTaskDef, Description, and
TimeoutSeconds properties; WorkflowTaskCoordinator accepts optional IMetadataClient
and batch-registers annotated task defs on startup; new
AddConductorWorkerWithMetadata DI helper wires everything up
Mirrors Python SDK fix (PRs #387 and #388). When the server returns 404 (endpoint not found) or 405 (Method Not Allowed on older Conductor 4.x), _useUpdateV2 is set to false permanently so v2 is never retried. Subsequent calls go directly to v1 without probing the server again. Also adds InternalsVisibleTo for test project and upgrades test TFM to net10.0. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ecLog in WorkflowTaskExecutor - Add NonRetryableException class: thrown by workers to mark task as FAILED_WITH_TERMINAL_ERROR with no retries - Wire ConductorMetrics counters/histograms at every poll, execute, and update-task path (was dead code) - Wire EventDispatcher.Instance at every key lifecycle event (OnPolling, OnPollSuccess, OnPollEmpty, OnPollError, OnTaskExecutionStarted, OnTaskExecutionCompleted, OnTaskExecutionFailed, OnTaskUpdateSent, OnTaskUpdateFailed) - Catch NonRetryableException separately and set FAILEDWITHTERMINALERROR status - Append TaskExecLog with full stack trace (e.ToString()) on both NonRetryable and regular failures Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.