groundswell 0.0.2 → 1.0.0

package/plan/001_d3bb02af4886/bugfix/architecture/concurrent_execution_best_practices.md

@@ -1,1565 +0,0 @@
-# Concurrent Workflow Execution and Error Handling: Best Practices Research
-
-**Author**: Research Specialist
-**Date**: 2026-01-12
-**Status**: Comprehensive Research Report
-**Target**: Groundswell Hierarchical Workflow Engine
-
----
-
-## Executive Summary
-
-This document provides comprehensive research findings on concurrent workflow execution and error handling patterns, with specific recommendations for the Groundswell workflow engine. The research covers Promise.all vs Promise.allSettled trade-offs, production-grade workflow engine patterns, error aggregation strategies, and fail-fast vs complete-all strategies.
-
-**Key Finding**: The current implementation uses `Promise.all()` in the @Task decorator with `concurrent: true`, which implements a fail-fast strategy. Based on PRD requirements and production patterns, this report recommends adding support for `Promise.allSettled()` to enable "complete all" error handling with proper error aggregation.
-
----
-
-## Table of Contents
-
-1. [Promise.all vs Promise.allSettled: Technical Comparison](#1-promiseall-vs-promiseallsettled-technical-comparison)
-2. [Production Workflow Engine Patterns](#2-production-workflow-engine-patterns)
-3. [Error Aggregation Strategies](#3-error-aggregation-strategies)
-4. [Fail-Fast vs Complete-All Decision Framework](#4-fail-fast-vs-complete-all-decision-framework)
-5. [Groundswell-Specific Recommendations](#5-groundswell-specific-recommendations)
-6. [Implementation Roadmap](#6-implementation-roadmap)
-7. [Appendix: Code Examples](#7-appendix-code-examples)
-
----
-
-## 1. Promise.all vs Promise.allSettled: Technical Comparison
-
-### 1.1 Behavioral Differences
-
-#### Promise.all (Fail-Fast Strategy)
-
-```typescript
-// Current Groundswell Implementation (src/decorators/task.ts:112)
-await Promise.all(runnable.map((w) => w.run()));
-
-// Behavior:
-// - Rejects IMMEDIATELY when ANY promise rejects
-// - Does NOT wait for other promises to complete
-// - Returns array of resolved values when ALL fulfill
-// - Use case: All operations must succeed, fast-fail desired
-```
-
-**Characteristics:**
-- **Performance**: Faster failure detection (stops at first error)
-- **Resource Usage**: Wastes resources from in-flight operations
-- **Error Visibility**: Only shows first error (hides concurrent failures)
-- **Debugging**: Difficult to understand full scope of failures
-
-**Example:**
-```typescript
-const results = await Promise.all([
-  task1(), // completes in 100ms
-  task2(), // completes in 200ms
-  task3(), // FAILS in 50ms
-]);
-// Execution stops at 50ms, results from task1/task2 lost
-// Only error from task3 is visible
-```
-
-#### Promise.allSettled (Complete-All Strategy)
-
-```typescript
-// Recommended Enhancement for Groundswell
-const results = await Promise.allSettled(
-  runnable.map((w) => w.run())
-);
-
-// Behavior:
-// - Waits for ALL promises to settle (fulfilled OR rejected)
-// - Returns array of status objects with reason/value
-// - Never rejects (always fulfills)
-// - Use case: Need all results regardless of individual failures
-```
-
-**Characteristics:**
-- **Performance**: Slower failure detection (waits for all operations)
-- **Resource Usage**: Completes all work (no waste)
-- **Error Visibility**: Shows ALL errors (complete picture)
-- **Debugging**: Easy to understand scope and patterns of failures
-
-**Example:**
-```typescript
-const results = await Promise.allSettled([
-  task1(), // completes in 100ms, fulfills
-  task2(), // completes in 200ms, fulfills
-  task3(), // FAILS in 50ms, rejects
-]);
-// Execution completes at 200ms (waits for slowest)
-// All three results available:
-// [
-//   { status: 'fulfilled', value: 'task1-result' },
-//   { status: 'fulfilled', value: 'task2-result' },
-//   { status: 'rejected', reason: Error('task3 failed') }
-// ]
-```
-
-### 1.2 Performance Comparison
-
-| Metric | Promise.all | Promise.allSettled |
-|--------|-------------|-------------------|
-| **Time to First Error** | Immediate (min time) | Waits for all (max time) |
-| **Time to Complete All** | N/A (fails early) | Max of all operations |
-| **Resource Efficiency** | Low (wastes in-flight work) | High (completes all work) |
-| **Error Completeness** | First error only | All errors |
-| **Partial Results** | Lost | Preserved |
-| **Backpressure** | Poor (can't throttle) | Better (can process as they settle) |
-
-### 1.3 When to Use Each Strategy
-
-#### Use Promise.all When:
-
-1. **All Operations Are Critical**: Partial success is meaningless
-   ```typescript
-   // Financial transaction - all parts must succeed
-   await Promise.all([
-     validateAccount(),
-     holdFunds(),
-     processTransfer(),
-     confirmReceipt(),
-   ]);
-   // If any fails, entire transaction should fail
-   ```
-
-2. **Fast Failure Saves Resources**: Stopping early prevents waste
-   ```typescript
-   // Batch API calls with early validation
-   const isValid = await validateBatch(items);
-   if (!isValid) {
-     return; // Don't process if validation fails
-   }
-   await Promise.all(items.map(processItem));
-   ```
-
-3. **Dependent Operations**: Subsequent steps depend on all succeeding
-   ```typescript
-   const [config, schema] = await Promise.all([
-     fetchConfig(),
-     fetchSchema(),
-   ]);
-   // Both needed for next step
-   validateData(config, schema);
-   ```
-
-4. **Error Cascading Acceptable**: First error is representative
-   ```typescript
-   // Homogeneous operations (same type of task)
-   await Promise.all(
-     servers.map(s => s.ping())
-   );
-   // If any server is down, treat all as down
-   ```
-
-#### Use Promise.allSettled When:
-
-1. **Partial Success Has Value**: Some results better than none
-   ```typescript
-   // Bulk data synchronization
-   const results = await Promise.allSettled(
-     records.map(r => syncRecord(r))
-   );
-   const succeeded = results.filter(r => r.status === 'fulfilled');
-   const failed = results.filter(r => r.status === 'rejected');
-   // Process succeeded, retry failed
-   ```
-
-2. **Independent Operations**: Tasks don't depend on each other
-   ```typescript
-   // Parallel notifications
-   await Promise.allSettled([
-     sendEmail(user),
-     sendSlack(user),
-     sendSMS(user),
-   ]);
-   // Try all channels, log failures, continue
-   ```
-
-3. **Error Aggregation Needed**: Need to see all failures
-   ```typescript
-   // Batch validation with comprehensive error reporting
-   const validations = await Promise.allSettled(
-     fields.map(f => validateField(f))
-   );
-   const errors = validations
-     .filter(v => v.status === 'rejected')
-     .map(v => v.reason);
-   // Return all validation errors to user
-   ```
-
-4. **Idempotent or Retryable Operations**: Can handle/ignore failures
-   ```typescript
-   // Cache warming
-   await Promise.allSettled(
-     keys.map(k => cache.set(k, fetchValue(k)))
-   );
-   // Populate as much as possible, retry misses later
-   ```
-
-### 1.4 Recommendations for Groundswell
-
-**Current State Analysis:**
-- File: `/home/dustin/projects/groundswell/src/decorators/task.ts`
-- Line 112: Uses `Promise.all()` for concurrent execution
-- Behavior: Fail-fast (stops at first child workflow error)
-
-**Issue**: This prevents partial success scenarios and hides concurrent errors.
-
-**Recommendation**: Support both strategies via configuration:
-
-```typescript
-export interface TaskOptions {
-  name?: string;
-  concurrent?: boolean;
-
-  // NEW: Add error handling strategy
-  errorStrategy?: 'fail-fast' | 'complete-all';
-
-  // NEW: Enable error aggregation
-  mergeErrors?: boolean;
-}
-```
-
----
-
-## 2. Production Workflow Engine Patterns
-
-### 2.1 Temporal.io: Child Workflow Error Handling
-
-**Pattern**: Temporal uses a "parent close policy" to determine child behavior on parent failure.
-
-**Key Concepts**:
-1. **Parent Close Policy**: What happens to children when parent completes
-2. **Cancellation Scopes**: Groups of activities that cancel together
-3. **Activity Retry Policies**: Automatic retries with exponential backoff
-
-**Temporal's Approach**:
-```typescript
-// Temporal-style child workflow execution
-const childHandle = await workflow.executeChild(childWorkflow, {
-  taskQueue: 'my-task-queue',
-  retryPolicy: {
-    initialInterval: '1 second',
-    backoffCoefficient: 2,
-    maximumAttempts: 5,
-    nonRetryableErrorTypes: ['PermanentError'],
-  },
-});
-
-// Children continue by default when parent completes
-// Can be overridden with:
-const options = {
-  parentClosePolicy: ParentClosePolicy.PARENT_CLOSE_POLICY_ABANDON,
-  // or REQUEST_CANCEL, TERMINATE
-};
-```
-
-**Lessons for Groundswell**:
-1. **Configurable Child Behavior**: Allow parent to specify what happens on error
-2. **Retry Policies**: Built-in retry with exponential backoff
-3. **Error Categorization**: Distinguish retryable vs non-retryable errors
-4. **Cancellation Propagation**: Controlled cascading of failures
-
-### 2.2 AWS Step Functions: Error Handling
-
-**Pattern**: Step Functions uses catch/cascade patterns with automatic retries.
-
-**Key Concepts**:
-1. **Retry Blocks**: Automatic retry with backoff
-2. **Catch Blocks**: Error handling and fallback
-3. **Parallel States**: Branch and merge with error aggregation
-
-**Step Functions Approach**:
-```json
-{
-  "Type": "Parallel",
-  "Branches": [
-    {
-      "StartAt": "ProcessA",
-      "States": {
-        "ProcessA": {
-          "Type": "Task",
-          "Retry": [
-            {
-              "ErrorEquals": ["States.TaskFailed"],
-              "IntervalSeconds": 1,
-              "MaxAttempts": 3,
-              "BackoffRate": 2.0
-            }
-          ],
-          "Catch": [
-            {
-              "ErrorEquals": ["States.ALL"],
-              "ResultPath": "$.error",
-              "Next": "FallbackA"
-            }
-          ]
-        }
-      }
-    }
-  ]
-}
-```
-
-**Lessons for Groundswell**:
-1. **Declarative Retry**: Configuration-driven retry logic
-2. **Error Categories**: Different handling for different error types
-3. **Fallback States**: Degraded but functional behavior on failure
-4. **Parallel Error Aggregation**: Collect errors from all branches
-
-### 2.3 Cadence Workflow: Error Propagation
-
-**Pattern**: Cadence uses exception-based error propagation with child workflow isolation.
-
-**Key Concepts**:
-1. **Workflow Exceptions**: Errors propagate up the tree
-2. **Child Workflow Isolation**: Child errors don't crash parent unless thrown
-3. **Timeout Handling**: Distinction between timeout and failure
-
-**Cadence Approach**:
-```go
-// Cadence-style child workflow
-childFuture, _ := workflow.ExecuteChildWorkflow(ctx, childWorkflow, input)
-
-var result string
-if err := childFuture.Get(ctx, &result); err != nil {
-  // Child failed, but parent can handle
-  if activities.IsTimeoutError(err) {
-    // Handle timeout specifically
-  } else if activities.IsCanceledError(err) {
-    // Handle cancellation
-  } else {
-    // Handle other errors
-  }
-  // Parent continues
-}
-```
-
-**Lessons for Groundswell**:
-1. **Non-Blocking Errors**: Parent can handle child failures
-2. **Error Type Discrimination**: Different error types handled differently
-3. **Explicit Error Handling**: Parent must explicitly handle child errors
-4. **Timeout as First-Class Concept**: Distinguish timeout from failure
-
-### 2.4 Apache Airflow: Task Failure Handling
-
-**Pattern**: Airflow uses DAG-based execution with trigger rules controlling downstream behavior.
-
-**Key Concepts**:
-1. **Trigger Rules**: Determine when tasks run based on upstream state
-2. **Task Instances**: Individual executions with state tracking
-3. **Retries**: Configurable per task
-4. **Failure Handlers**: Callbacks on task failure
-
-**Airflow Approach**:
-```python
-# Airflow-style DAG with trigger rules
-task1 = PythonOperator(
-    task_id='task1',
-    python_callable=process_data,
-    retries=3,
-    retry_delay=timedelta(seconds=60),
-    on_failure_callback=notify_failure,
-)
-
-task2 = PythonOperator(
-    task_id='task2',
-    python_callable=process_more_data,
-    trigger_rule=TriggerRule.ONE_SUCCESS,  # Run if ANY upstream succeeded
-    depends_on_past=False,
-)
-
-task3 = PythonOperator(
-    task_id='task3',
-    python_callable=finalize,
-    trigger_rule=TriggerRule.ALL_DONE,  # Run after ALL upstream complete (success or fail)
-)
-```
-
-**Lessons for Groundswell**:
-1. **Flexible Trigger Rules**: Control downstream execution based on upstream state
-2. **Per-Task Retry Configuration**: Fine-grained retry control
-3. **Failure Callbacks**: Reactive error handling
-4. **Completion-Based Triggers**: "ALL_DONE" = Promise.allSettled pattern
-
-### 2.5 Cross-Cutting Patterns Summary
-
-| Pattern | Temporal | Step Functions | Cadence | Airflow | Groundswell |
-|---------|----------|----------------|---------|---------|-------------|
-| **Child Isolation** | ✓ | ✓ | ✓ | ✓ | ⚠️ Partial |
-| **Error Aggregation** | ✗ | ✓ (Parallel) | ⚠️ Manual | ✓ (Trigger rules) | ✗ |
-| **Automatic Retry** | ✓ | ✓ | ✓ | ✓ | ✗ |
-| **Fail-Fast** | Default | Configurable | Default | Configurable | Default |
-| **Complete-All** | ✗ | ✓ (Parallel) | ⚠️ Manual | ✓ (ALL_DONE) | ✗ |
-| **Fallback States** | ✗ | ✓ (Catch) | ⚠️ Manual | ✓ (On failure) | ✗ |
-| **Cancellation Propagation** | ✓ | ✓ | ✓ | ✓ | ⚠️ Unimplemented |
-
----
-
-## 3. Error Aggregation Strategies
-
-### 3.1 Strategy 1: First Error Wins (Current Groundswell)
-
-**Description**: Return immediately on first error, hiding concurrent failures.
-
-**Implementation**:
-```typescript
-// Current implementation (src/decorators/task.ts:112)
-await Promise.all(runnable.map((w) => w.run()));
-```
-
-**Pros**:
-- Simple implementation
-- Fast failure detection
-- Low memory overhead
-- Familiar pattern (Promise.all)
-
-**Cons**:
-- Hides concurrent errors
-- Loses partial results
-- Difficult debugging
-- No error context
-
-**Use Case**: Critical workflows where partial success is meaningless.
-
-### 3.2 Strategy 2: Collect All Errors (Recommended)
-
-**Description**: Wait for all operations to complete, collect all errors.
-
-**Implementation**:
-```typescript
-// Promise.allSettled with error aggregation
-const results = await Promise.allSettled(
-  runnable.map((w) => w.run())
-);
-
-const errors = results
-  .filter((r) => r.status === 'rejected')
-  .map((r, idx) => ({
-    workflow: runnable[idx],
-    error: r.status === 'rejected' ? r.reason : undefined,
-  }));
-
-if (errors.length > 0) {
-  throw new AggregateError(
-    errors.map(e => e.error),
-    `${errors.length} child workflows failed`
-  );
-}
-```
-
-**Pros**:
-- Complete error visibility
-- Preserves partial results
-- Better debugging
-- Error pattern analysis
-
-**Cons**:
-- Higher memory usage
-- Slower failure detection
-- More complex error handling
-- Need AggregateError polyfill for older Node versions
-
-**Use Case**: Independent operations where understanding all failures is important.
-
-### 3.3 Strategy 3: Hierarchical Error Merging
-
-**Description**: Merge errors with workflow hierarchy context.
-
-**Implementation**:
-```typescript
-class WorkflowAggregateError extends Error {
-  constructor(
-    message: string,
-    public errors: WorkflowError[],
-    public workflowId: string,
-    public workflowName: string,
-    public parent?: WorkflowAggregateError
-  ) {
-    super(message);
-    this.name = 'WorkflowAggregateError';
-  }
-
-  // Pretty-print error hierarchy
-  toString(): string {
-    let output = `${this.workflowName}: ${this.message}\n`;
-    for (const error of this.errors) {
-      output += `  - ${error.workflowId}: ${error.message}\n`;
-    }
-    if (this.parent) {
-      output += `\nCaused by:\n${this.parent.toString()}`;
-    }
-    return output;
-  }
-
-  // Get error statistics
-  getStats() {
-    return {
-      totalErrors: this.errors.length,
-      byWorkflow: groupErrorsByWorkflow(this.errors),
-      byErrorType: groupErrorsByType(this.errors),
-    };
-  }
-}
-```
-
-**Pros**:
-- Maintains workflow context
-- Hierarchical error visualization
-- Rich error metadata
-- Statistical analysis
-
-**Cons**:
-- Complex implementation
-- Higher memory overhead
-- Steeper learning curve
-
-**Use Case**: Complex workflow trees needing comprehensive error reporting.
-
-### 3.4 Strategy 4: Error Rate Thresholding
-
-**Description**: Allow some failures, only throw if error rate exceeds threshold.
-
-**Implementation**:
-```typescript
-interface ErrorThresholdStrategy {
-  enabled: boolean;
-  maxErrorRate: number; // 0.0 to 1.0
-  minAbsoluteErrors: number; // Throw if >= this many errors
-}
-
-async function executeWithThreshold(
-  workflows: Workflow[],
-  strategy: ErrorThresholdStrategy
-) {
-  const results = await Promise.allSettled(
-    workflows.map((w) => w.run())
-  );
-
-  const failures = results.filter((r) => r.status === 'rejected');
-  const errorRate = failures.length / results.length;
-
-  const shouldThrow =
-    failures.length >= strategy.minAbsoluteErrors ||
-    errorRate > strategy.maxErrorRate;
-
-  if (shouldThrow) {
-    throw new WorkflowAggregateError(
-      `${failures.length}/${results.length} workflows failed (${(errorRate * 100).toFixed(1)}%)`,
-      failures.map(f => f.reason),
-      // ... context
-    );
-  }
-
-  return results;
-}
-```
-
-**Pros**:
-- Graceful degradation
-- Configurable tolerance
-- Good for bulk operations
-- Prevents single failures from stopping everything
-
-**Cons**:
-- May hide systematic issues
-- Requires threshold tuning
-- Ambiguous success criteria
-
-**Use Case**: High-volume operations where some failures are acceptable (e.g., bulk notifications, cache warming).
-
-### 3.5 Error Aggregation Best Practices
-
-#### DO:
-1. **Preserve Workflow Context**: Include workflow ID, name, and path in errors
-2. **Maintain Error Order**: Keep errors in execution order for debugging
-3. **Include Error Metadata**: Timestamps, retry counts, state snapshots
-4. **Provide Statistics**: Error counts, rates, patterns
-5. **Support Error Inspection**: Allow programmatic error analysis
-
-#### DON'T:
-1. **Lose Stack Traces**: Always preserve original stack traces
-2. **Swallow Errors**: Always report or log errors
-3. **Create Excessive Depth**: Limit error nesting to prevent stack overflow
-4. **Duplicate Errors**: Deduplicate identical errors
-5. **Include Sensitive Data**: Redact passwords, tokens, PII from errors
-
-### 3.6 Groundswell Implementation Recommendation
-
-**Current State**:
-- Type: `/home/dustin/projects/groundswell/src/types/error-strategy.ts`
-- Status: Defined but not implemented
-- Interface: `ErrorMergeStrategy` exists but unused
-
-**Recommendation**:
-
-```typescript
-// Enhanced error strategy interface
-export interface ErrorStrategy {
-  type: 'fail-fast' | 'complete-all' | 'threshold';
-  threshold?: number; // For 'threshold' type
-  mergeErrors?: boolean; // Always merge errors in 'complete-all' mode
-}
-
-// Implementation in @Task decorator
-if (opts.concurrent && Array.isArray(result)) {
-  const runnable = workflows.filter(/* ... */);
-
-  if (runnable.length > 0) {
-    const errorStrategy = opts.errorStrategy || 'fail-fast';
-
-    if (errorStrategy === 'fail-fast') {
-      // Current behavior: Promise.all
-      await Promise.all(runnable.map((w) => w.run()));
-    } else {
-      // New behavior: Promise.allSettled with aggregation
-      const results = await Promise.allSettled(
-        runnable.map((w) => w.run())
-      );
-
-      const errors = results
-        .filter((r) => r.status === 'rejected')
-        .map((r) => r.reason);
-
-      if (errors.length > 0) {
-        throw new WorkflowAggregateError(
-          `${errors.length} child workflows failed`,
-          errors,
-          wf.id,
-          taskName
-        );
-      }
-    }
-  }
-}
-```
-
----
-
-## 4. Fail-Fast vs Complete-All Decision Framework
-
-### 4.1 Decision Tree
-
-```
-                    Start Concurrent Tasks
-                              |
-                    Are tasks independent?
-                     /              \
-                   No                Yes
-                   |                  |
-            Use Promise.all      Is partial success
-             (Fail-Fast)         valuable?
-                                   /            \
-                                 No              Yes
-                                 |                |
-                          Use Promise.all    Need complete
-                           (Fail-Fast)       error picture?
-                                             /          \
-                                           No          Yes
-                                           |             |
-                                    Use Promise    Use Promise
-                                    .allSettled   .allSettled
-                                    (simple)      (aggregate)
-```
-
-### 4.2 Heuristics for Choosing Strategy
-
-#### Use Fail-Fast (Promise.all) When:
-
-1. **Strong Coupling**: Tasks are logically dependent
-   - Example: Multi-step transaction where all steps must succeed
-   - Pattern: `validate → hold → transfer → confirm`
-
-2. **Fast Failure Saves Money**: Early termination prevents waste
-   - Example: Paid API calls with rate limits
-   - Pattern: Stop if quota exceeded or auth fails
-
-3. **First Error Is Representative**: Errors are correlated
-   - Example: All tasks use same resource (database, API)
-   - Pattern: If one fails, all would fail
-
-4. **User Experience Requires Speed**: Fast feedback is critical
-   - Example: Form validation before submission
-   - Pattern: Show first error immediately
-
-#### Use Complete-All (Promise.allSettled) When:
-
-1. **Independent Tasks**: No dependencies between operations
-   - Example: Sending notifications via multiple channels
-   - Pattern: Email, SMS, Slack - try all, report failures
-
-2. **Partial Success Has Value**: Some results better than none
-   - Example: Bulk data synchronization
-   - Pattern: Sync what succeeds, retry failures
-
-3. **Error Analysis Required**: Need to see all failures
-   - Example: Batch processing with quality control
-   - Pattern: Analyze error patterns, adjust strategy
-
-4. **Idempotent Operations**: Can safely retry failures
-   - Example: Cache warming, data replication
-   - Pattern: Populate cache, retry misses later
-
-### 4.3 Strategy Selection Matrix
-
-| Scenario | Fail-Fast | Complete-All | Rationale |
-|----------|-----------|--------------|-----------|
-| **Financial Transaction** | ✓ | ✗ | All-or-nothing, can't have partial money movement |
-| **Bulk Notifications** | ✗ | ✓ | Try all channels, log failures |
-| **Data Validation** | ✓ | ✗ | Fast feedback, first error representative |
-| **Batch Processing** | ✗ | ✓ | Process what you can, retry failures |
-| **Cache Warming** | ✗ | ✓ | Populate as much as possible |
-| **API Rate Limiting** | ✓ | ✗ | Stop immediately to avoid waste |
-| **Multi-Region Deployment** | ✗ | ✓ | Deploy to available regions, log failures |
-| **Form Submission** | ✓ | ✗ | User needs immediate feedback |
-| **Data Replication** | ✗ | ✓ | Replicate to available nodes |
-| **Paid API Calls** | ✓ | ✗ | Save money on failed requests |
-
-### 4.4 Hybrid Strategy: Adaptive Error Handling
-
-**Concept**: Start with complete-all, switch to fail-fast if error rate exceeds threshold.
-
-**Implementation**:
-```typescript
-async function adaptiveExecute(
-  workflows: Workflow[],
-  options: {
-    initialStrategy: 'fail-fast' | 'complete-all';
-    errorRateThreshold: number; // Switch to fail-fast if exceeded
-    minSampleSize: number; // Don't switch until this many tasks
-  }
-) {
-  let strategy = options.initialStrategy;
-  let completed = 0;
-  let errors = 0;
-
-  const results = await Promise.allSettled(
-    workflows.map(async (workflow) => {
-      try {
-        const result = await workflow.run();
-        completed++;
-        errors++;
-
-        // Check if we should switch to fail-fast
-        if (
-          strategy === 'complete-all' &&
-          completed >= options.minSampleSize
-        ) {
-          const errorRate = errors / completed;
-          if (errorRate > options.errorRateThreshold) {
-            // Switch to fail-fast
-            throw new Error(`Error rate ${(errorRate * 100).toFixed(1)}% exceeds threshold, switching to fail-fast`);
-          }
-        }
-
-        return { status: 'fulfilled', value: result };
-      } catch (error) {
-        return { status: 'rejected', reason: error };
-      }
-    })
-  );
-
-  return results;
-}
-```
-
-**Use Case**: High-volume operations where most tasks succeed, but rapid failure indicates systemic issues.
-
-### 4.5 Groundswell Decision Framework
-
-**Question**: Should Groundswell use fail-fast or complete-all by default?
-
-**Analysis**:
-
-**Arguments for Fail-Fast (Current)**:
-- Workflow engines traditionally emphasize correctness over partial success
-- Parent workflows often need to know immediately if children fail
-- Simpler mental model for developers
-- Lower memory overhead
-- Already implemented
-
-**Arguments for Complete-All (Recommended)**:
-- Observability is a core feature (PRD emphasis on logging and debugging)
-- Child workflows are often independent (parallel processing pattern)
-- "Complete all" provides better error visibility
-- Aligns with production patterns (Airflow, Step Functions)
-- Enables graceful degradation patterns
-- More flexible (can opt-in to fail-fast)
-
-**Recommendation**: **Default to complete-all with opt-in fail-fast**
-
-**Rationale**:
-1. Observability First: Groundswell's value prop is debugging and observability
-2. Independent by Default: @Task creates independent child workflows
-3. Error Visibility: Complete-all provides comprehensive error information
-4. Production Patterns: Aligns with Airflow and Step Functions
-5. Flexibility: Developers can opt-in to fail-fast when needed
-
----
-
-## 5. Groundswell-Specific Recommendations
-
-### 5.1 Current Implementation Analysis
-
-**File**: `/home/dustin/projects/groundswell/src/decorators/task.ts`
-
-**Current Behavior (Line 104-114)**:
-```typescript
-// If concurrent option is set and we have multiple workflows, run them in parallel
-if (opts.concurrent && Array.isArray(result)) {
-  const runnable = workflows.filter(
-    (w): w is WorkflowClass =>
-      w && typeof w === 'object' && 'run' in w && typeof w.run === 'function'
-  );
-
-  if (runnable.length > 0) {
-    await Promise.all(runnable.map((w) => w.run()));
-  }
-}
-```
-
-**Characteristics**:
-- Strategy: Fail-fast (Promise.all)
-- Error Handling: First error wins
-- Partial Results: Lost
-- Error Aggregation: None
-- Configuration: None (hardcoded)
-
-**Issues**:
-1. No error aggregation: Only first error is visible
-2. No flexibility: Can't choose strategy
-3. No observability: Loses error information
-4. No graceful degradation: All-or-nothing approach
-
-### 5.2 Recommended Enhancement
-
-#### Phase 1: Add Error Strategy Option
-
-**Changes to `/home/dustin/projects/groundswell/src/types/decorators.ts`**:
-
-```typescript
-export interface TaskOptions {
-  name?: string;
-  concurrent?: boolean;
-
-  // NEW: Error handling strategy
-  errorStrategy?: 'fail-fast' | 'complete-all';
-
-  // NEW: Error aggregation (only for 'complete-all')
-  mergeErrors?: boolean;
-}
-```
-
-**Changes to `/home/dustin/projects/groundswell/src/decorators/task.ts`**:
-
-```typescript
-// If concurrent option is set and we have multiple workflows, run them in parallel
-if (opts.concurrent && Array.isArray(result)) {
-  const runnable = workflows.filter(
-    (w): w is WorkflowClass =>
-      w && typeof w === 'object' && 'run' in w && typeof w.run === 'function'
-  );
-
-  if (runnable.length > 0) {
-    const strategy = opts.errorStrategy || 'fail-fast';
-
-    if (strategy === 'fail-fast') {
-      // Current behavior: Promise.all (fail-fast)
-      await Promise.all(runnable.map((w) => w.run()));
-    } else {
-      // New behavior: Promise.allSettled (complete-all)
-      const settledResults = await Promise.allSettled(
-        runnable.map((w) => w.run())
-      );
-
-      // Collect errors
-      const errors = settledResults
-        .map((result, idx) => ({ result, workflow: runnable[idx] }))
-        .filter(({ result }) => result.status === 'rejected')
-        .map(({ result, workflow }) => ({
-          workflowId: workflow.id,
-          workflowName: workflow.constructor.name,
-          error: result.status === 'rejected' ? result.reason : undefined,
-        }));
-
-      // Throw if there were errors
-      if (errors.length > 0) {
-        // Create aggregate error with all failures
-        const aggregateError = new Error(
-          `${errors.length} child workflow(s) failed in task '${taskName}'`
-        ) as any;
-
-        aggregateError.name = 'WorkflowAggregateError';
-        aggregateError.errors = errors;
-        aggregateError.taskName = taskName;
-        aggregateError.workflowId = wf.id;
-        aggregateError.totalChildren = runnable.length;
-        aggregateError.failedChildren = errors.length;
-
-        throw aggregateError;
-      }
-    }
-  }
-}
-```
-
-#### Phase 2: Implement AggregateError Type
-
-**New file**: `/home/dustin/projects/groundswell/src/types/aggregate-error.ts`
-
-```typescript
-import type { WorkflowError } from './error.js';
-
-/**
- * Aggregate error containing multiple child workflow errors
- */
-export interface WorkflowAggregateError extends Error {
-  name: 'WorkflowAggregateError';
-  message: string;
-  errors: Array<{
-    workflowId: string;
-    workflowName: string;
-    error: unknown;
-  }>;
-  taskName: string;
-  workflowId: string;
-  totalChildren: number;
-  failedChildren: number;
-}
-
-/**
- * Type guard for WorkflowAggregateError
- */
-export function isWorkflowAggregateError(
-  error: unknown
-): error is WorkflowAggregateError {
-  return (
-    error instanceof Error &&
-    (error as any).name === 'WorkflowAggregateError' &&
-    'errors' in error &&
-    'totalChildren' in error &&
-    'failedChildren' in error
-  );
-}
-
-/**
- * Create a workflow aggregate error
- */
-export function createAggregateError(
-  errors: Array<{ workflowId: string; workflowName: string; error: unknown }>,
-  taskName: string,
-  parentWorkflowId: string
-): WorkflowAggregateError {
-  const error = new Error(
-    `${errors.length} child workflow(s) failed in task '${taskName}'`
-  ) as WorkflowAggregateError;
-
-  error.name = 'WorkflowAggregateError';
-  error.errors = errors;
-  error.taskName = taskName;
-  error.workflowId = parentWorkflowId;
-  error.totalChildren = errors.length; // Will be updated by caller
-  error.failedChildren = errors.length;
-
-  return error;
-}
-```
-
-#### Phase 3: Update Error Handling Documentation
-
-**File**: `/home/dustin/projects/groundswell/docs/workflow.md`
-
-**Add section**:
-
-```markdown
-## Error Handling in Concurrent Tasks
-
-### Fail-Fast Strategy (Default)
-
-When `concurrent: true` is used without specifying `errorStrategy`, the workflow uses a fail-fast approach:
-
-```typescript
-class ParentWorkflow extends Workflow {
-  @Task({ concurrent: true })
-  async createChildren(): Promise<ChildWorkflow[]> {
-    return [
-      new ChildWorkflow('child1', this),
-      new ChildWorkflow('child2', this),
-      new ChildWorkflow('child3', this),
-    ];
-  }
-
-  async run() {
-    try {
-      await this.createChildren();
-    } catch (error) {
-      // Throws immediately when first child fails
-      console.error('First child failed:', error.message);
-    }
-  }
-}
-```
-
-### Complete-All Strategy
-
-To wait for all children to complete and collect all errors:
-
-```typescript
-class ParentWorkflow extends Workflow {
-  @Task({
-    concurrent: true,
-    errorStrategy: 'complete-all', // Wait for all children
-  })
-  async createChildren(): Promise<ChildWorkflow[]> {
-    return [
-      new ChildWorkflow('child1', this),
-      new ChildWorkflow('child2', this),
-      new ChildWorkflow('child3', this),
-    ];
-  }
-
-  async run() {
-    try {
-      await this.createChildren();
-    } catch (error) {
-      if (isWorkflowAggregateError(error)) {
-        console.error(`${error.failedChildren}/${error.totalChildren} children failed:`);
-
-        for (const failure of error.errors) {
-          console.error(`  - ${failure.workflowName}: ${failure.error?.message}`);
-        }
-      }
-    }
-  }
-}
-```
-
-### Choosing the Right Strategy
-
-**Use Fail-Fast When:**
-- Child workflows are dependent on each other
-- Partial success is not meaningful
-- You need immediate feedback on errors
-- Resource conservation is important
-
-**Use Complete-All When:**
-- Child workflows are independent
-- Partial results have value
-- You need comprehensive error reporting
-- Observability is a priority
-```
-
-### 5.3 Backward Compatibility
-
-**Recommendation**: Keep current behavior (fail-fast) as default to maintain backward compatibility.
-
-**Rationale**:
-1. Existing code won't break
-2. Opt-in to new behavior via `errorStrategy: 'complete-all'`
-3. Gradual migration path
-4. Safe default (fail-fast is more conservative)
-
-**Future Enhancement**: Consider changing default to `complete-all` in a major version (2.0).
-
-### 5.4 Testing Recommendations
-
-**Add tests to** `/home/dustin/projects/groundswell/src/__tests__/adversarial/edge-case.test.ts`:
-
-```typescript
-describe('Concurrent Error Handling', () => {
-  it('should use fail-fast strategy by default', async () => {
-    // Test that current behavior is maintained
-  });
-
-  it('should use complete-all strategy when specified', async () => {
-    // Test new behavior with errorStrategy: 'complete-all'
-  });
-
-  it('should aggregate all errors in complete-all mode', async () => {
-    // Test that all errors are collected
-  });
-
-  it('should preserve error context in aggregate error', async () => {
-    // Test that workflow ID, name, etc. are preserved
-  });
-
-  it('should throw WorkflowAggregateError in complete-all mode', async () => {
-    // Test error type and structure
-  });
-
-  it('should handle partial success in complete-all mode', async () => {
-    // Test that some successes + some failures works correctly
-  });
-});
-```
-
----
-
-## 6. Implementation Roadmap
-
-### Phase 1: Core Implementation (Week 1)
-
-**Tasks**:
-1. Add `errorStrategy` option to `TaskOptions` interface
-2. Implement `Promise.allSettled` logic in @Task decorator
-3. Create `WorkflowAggregateError` type and factory
-4. Add error collection and aggregation logic
-5. Update JSDoc comments
-
-**Deliverables**:
-- Enhanced `/home/dustin/projects/groundswell/src/types/decorators.ts`
-- Updated `/home/dustin/projects/groundswell/src/decorators/task.ts`
-- New `/home/dustin/projects/groundswell/src/types/aggregate-error.ts`
-
-**Success Criteria**:
-- TypeScript compiles without errors
-- Existing tests pass (backward compatibility)
-- New behavior accessible via `errorStrategy: 'complete-all'`
-
-### Phase 2: Testing (Week 1-2)
-
-**Tasks**:
-1. Add unit tests for error aggregation
-2. Add integration tests for complete-all strategy
-3. Add edge case tests (mixed success/failure)
-4. Add performance tests (compare fail-fast vs complete-all)
-5. Update example workflows
-
-**Deliverables**:
-- New tests in `/home/dustin/projects/groundswell/src/__tests__/adversarial/edge-case.test.ts`
-- Updated `/home/dustin/projects/groundswell/examples/examples/06-concurrent-tasks.ts`
-
-**Success Criteria**:
-- All tests pass (including new tests)
-- Test coverage > 90% for new code
-- Examples demonstrate both strategies
-
-### Phase 3: Documentation (Week 2)
-
-**Tasks**:
-1. Update `/home/dustin/projects/groundswell/docs/workflow.md`
-2. Add error handling guide
-3. Update API reference
-4. Create migration guide
-5. Add best practices document
-
-**Deliverables**:
-- Updated documentation
-- New error handling guide
-- Migration guide from fail-fast to complete-all
-- Best practices document
-
-**Success Criteria**:
-- All APIs documented
-- Examples cover both strategies
-- Migration path is clear
-
-### Phase 4: Validation (Week 2)
-
-**Tasks**:
-1. Manual testing with real workflows
-2. Performance benchmarking
-3. Memory profiling
-4. Error scenario testing
-5. User acceptance testing
-
-**Deliverables**:
-- Performance benchmarks
-- Memory profiles
-- Test results report
-- Bug fixes (if any)
-
-**Success Criteria**:
-- Performance acceptable (< 10% overhead)
-- No memory leaks
-- All error scenarios handled correctly
-- Users can use both strategies effectively
-
----
-
-## 7. Appendix: Code Examples
-
-### 7.1 Complete Example: Fail-Fast vs Complete-All
-
-```typescript
-import { Workflow, Task, Step } from 'groundswell';
-
-// Child workflow that may fail
-class DataProcessor extends Workflow {
-  constructor(
-    public id: string,
-    public shouldFail: boolean = false,
-    parent?: Workflow
-  ) {
-    super(id, parent);
-  }
-
-  @Step()
-  async process(): Promise<string> {
-    if (this.shouldFail) {
-      throw new Error(`Processing failed in ${this.id}`);
-    }
-    return `Processed by ${this.id}`;
-  }
-
-  async run(): Promise<string> {
-    this.setStatus('running');
-    const result = await this.process();
-    this.setStatus('completed');
-    return result;
-  }
-}
-
-// Parent workflow using fail-fast (default)
-class FailFastParent extends Workflow {
-  @Task({ concurrent: true })
-  async spawnProcessors(): Promise<DataProcessor[]> {
-    return [
-      new DataProcessor('processor-1', false, this),
-      new DataProcessor('processor-2', true, this),  // This will fail
-      new DataProcessor('processor-3', false, this),
-    ];
-  }
-
-  async run(): Promise<void> {
-    this.setStatus('running');
-
-    try {
-      await this.spawnProcessors();
-      console.log('All processors completed successfully');
-    } catch (error) {
-      console.error('Fail-fast: Stopped at first error');
-      console.error('Error:', error.message);
-      // processor-3 never runs
-      // Only processor-2's error is visible
-    }
-
-    this.setStatus('completed');
-  }
-}
-
-// Parent workflow using complete-all
-class CompleteAllParent extends Workflow {
-  @Task({
-    concurrent: true,
-    errorStrategy: 'complete-all',  // NEW: Wait for all
-  })
-  async spawnProcessors(): Promise<DataProcessor[]> {
-    return [
-      new DataProcessor('processor-1', false, this),
-      new DataProcessor('processor-2', true, this),  // This will fail
-      new DataProcessor('processor-3', false, this),
-    ];
-  }
-
-  async run(): Promise<void> {
-    this.setStatus('running');
-
-    try {
-      await this.spawnProcessors();
-      console.log('All processors completed successfully');
-    } catch (error) {
-      if (error.name === 'WorkflowAggregateError') {
-        console.error('Complete-all: All processors completed');
-        console.error(`${error.failedChildren}/${error.totalChildren} failed:`);
-
-        for (const failure of error.errors) {
-          console.error(`  - ${failure.workflowName}: ${failure.error?.message}`);
-        }
-
-        // processor-1 and processor-3 completed successfully
-        // processor-2's error is visible
-        // All errors are aggregated
-      }
-    }
-
-    this.setStatus('completed');
-  }
-}
-```
-
-### 7.2 Error Aggregation Implementation
-
-```typescript
-/**
- * Aggregate multiple workflow errors into a single error
- */
-class WorkflowAggregateError extends Error {
-  constructor(
-    message: string,
-    public errors: Array<{
-      workflowId: string;
-      workflowName: string;
-      error: unknown;
-    }>,
-    public taskName: string,
-    public workflowId: string,
-    public totalChildren: number,
-    public failedChildren: number
-  ) {
-    super(message);
-    this.name = 'WorkflowAggregateError';
-  }
-
-  /**
-   * Pretty-print the aggregate error
-   */
-  toString(): string {
-    let output = `${this.taskName}: ${this.message}\n`;
-    output += `  Total: ${this.totalChildren}, Failed: ${this.failedChildren}\n`;
-    output += `  Success Rate: ${((1 - this.failedChildren / this.totalChildren) * 100).toFixed(1)}%\n`;
-
-    for (const failure of this.errors) {
-      output += `  - ${failure.workflowName} (${failure.workflowId}):\n`;
-      output += `    ${failure.error?.message || 'Unknown error'}\n`;
-    }
-
-    return output;
-  }
-
-  /**
-   * Get error statistics
-   */
-  getStats() {
-    const errorsByType = new Map<string, number>();
-    const errorsByWorkflow = new Map<string, number>();
-
-    for (const failure of this.errors) {
-      const errorType = failure.error?.constructor.name || 'Unknown';
-      errorsByType.set(errorType, (errorsByType.get(errorType) || 0) + 1);
-      errorsByWorkflow.set(
-        failure.workflowName,
-        (errorsByWorkflow.get(failure.workflowName) || 0) + 1
-      );
-    }
-
-    return {
-      totalErrors: this.errors.length,
-      totalChildren: this.totalChildren,
-      failedChildren: this.failedChildren,
-      successRate: (1 - this.failedChildren / this.totalChildren) * 100,
-      errorsByType: Object.fromEntries(errorsByType),
-      errorsByWorkflow: Object.fromEntries(errorsByWorkflow),
-    };
-  }
-}
-```
-
-### 7.3 Usage in Production
-
-```typescript
-// Real-world example: Bulk data synchronization
-class DataSyncWorkflow extends Workflow {
-  @ObservedState()
-  recordsProcessed = 0;
-
-  @ObservedState()
-  recordsFailed = 0;
-
-  @ObservedState()
-  syncStartTime = 0;
-
-  @Task({
-    concurrent: true,
-    errorStrategy: 'complete-all',  // Try all records
-  })
-  async syncRecords(records: Record[]): Promise<RecordSyncWorkflow[]> {
-    return records.map(
-      (record) => new RecordSyncWorkflow(record, this)
-    );
-  }
-
-  @Step({ snapshotState: true })
-  async aggregateResults(results: PromiseSettledResult<unknown>[]): Promise<void> {
-    const succeeded = results.filter((r) => r.status === 'fulfilled');
-    const failed = results.filter((r) => r.status === 'rejected');
-
-    this.recordsProcessed = succeeded.length;
-    this.recordsFailed = failed.length;
-
-    this.logger.info(`Sync complete: ${succeeded.length} succeeded, ${failed.length} failed`);
-
-    if (failed.length > 0) {
-      this.logger.warn(`${failed.length} records failed to sync, will retry`);
-    }
-  }
-
-  async run(records: Record[]): Promise<SyncResults> {
-    this.setStatus('running');
-    this.syncStartTime = Date.now();
-
-    try {
-      await this.syncRecords(records);
-
-      // Continue despite partial failures
-      return {
-        total: records.length,
-        succeeded: this.recordsProcessed,
-        failed: this.recordsFailed,
-        duration: Date.now() - this.syncStartTime,
-      };
-    } catch (error) {
-      if (error.name === 'WorkflowAggregateError') {
-        // Log all failures but don't fail the workflow
-        this.logger.warn(`Partial sync complete: ${error.failedChildren}/${error.totalChildren} failed`);
-
-        return {
-          total: records.length,
-          succeeded: error.totalChildren - error.failedChildren,
-          failed: error.failedChildren,
-          duration: Date.now() - this.syncStartTime,
-          errors: error.errors,
-        };
-      }
-      throw error;
-    } finally {
-      this.setStatus('completed');
-    }
-  }
-}
-```
-
----
-
-## 8. Conclusion
-
-### 8.1 Summary of Findings
-
-1. **Promise.all vs Promise.allSettled**:
-   - Promise.all (fail-fast): Fast, simple, but loses error information
-   - Promise.allSettled (complete-all): Slower, more complex, but provides complete error visibility
-   - **Recommendation**: Support both strategies via configuration
-
-2. **Production Workflow Engines**:
-   - Temporal: Parent close policies, configurable cancellation
-   - Step Functions: Parallel states with error aggregation
-   - Cadence: Exception-based error propagation
-   - Airflow: Trigger rules for flexible execution
-   - **Gap**: Groundswell lacks error aggregation and flexible error strategies
-
-3. **Error Aggregation Strategies**:
-   - First error wins: Simple but limited (current implementation)
-   - Collect all errors: Comprehensive error visibility (recommended)
-   - Hierarchical merging: Maintains workflow context
-   - Threshold-based: Graceful degradation
-   - **Recommendation**: Implement collect-all with hierarchical context
-
-4. **Fail-Fast vs Complete-All**:
-   - Fail-fast: Critical workflows, dependent tasks, fast feedback
-   - Complete-all: Independent tasks, partial success, observability
-   - **Recommendation**: Default to complete-all for Groundswell (observability-first design)
-
-### 8.2 Recommendations for Groundswell
-
-**Priority 1 (Immediate)**:
-1. Add `errorStrategy` option to `TaskOptions`
-2. Implement `Promise.allSettled` path in @Task decorator
-3. Create `WorkflowAggregateError` type
-4. Maintain backward compatibility (fail-fast as default)
-
-**Priority 2 (Short-term)**:
-1. Add comprehensive error aggregation
-2. Include workflow context in errors
-3. Add error statistics and pretty-printing
-4. Update documentation and examples
-
-**Priority 3 (Long-term)**:
-1. Consider changing default to complete-all in v2.0
-2. Add retry policies per child workflow
-3. Implement threshold-based error handling
-4. Add error rate monitoring and alerting
-
-### 8.3 Impact Assessment
-
-**Benefits**:
-- Better observability: See all errors, not just first
-- Graceful degradation: Partial success scenarios
-- Production-ready: Aligns with industry patterns
-- Debugging: Comprehensive error information
-- Flexibility: Developers choose strategy per task
-
-**Costs**:
-- Complexity: More configuration options
-- Memory: Storing all errors (mitigated by error limiting)
-- Performance: Waiting for all tasks to complete (mitigated by configurable strategy)
-- Learning curve: Developers need to understand both strategies
-
-**Risk Mitigation**:
-- Backward compatibility: Default to fail-fast
-- Gradual migration: Opt-in to complete-all
-- Documentation: Clear examples and best practices
-- Testing: Comprehensive test coverage
-
----
-
-## References
-
-### Production Documentation
-
-1. **Temporal.io**: Child Workflows and Error Handling
-   - https://docs.temporal.io/develop/typescript/child-workflows
-   - Key concepts: Parent close policy, cancellation scopes, retry policies
-
-2. **AWS Step Functions**: Error Handling
-   - https://docs.aws.amazon.com/step-functions/latest/dg/concepts-error-handling.html
-   - Key concepts: Retry blocks, catch blocks, parallel state error aggregation
-
-3. **Cadence Workflow**: Execution Patterns
-   - https://cadenceworkflow.io/docs/concepts/execution/
-   - Key concepts: Exception-based error propagation, child workflow isolation
-
-4. **Apache Airflow**: Task Instances and Trigger Rules
-   - https://airflow.apache.org/docs/apache-airflow/stable/core-concepts/dags.html
-   - Key concepts: Trigger rules, task dependencies, failure handlers
-
-### JavaScript/TypeScript Documentation
-
-5. **MDN: Promise.all()**
-   - https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Promise/all
-   - Fail-fast behavior, rejection handling
-
-6. **MDN: Promise.allSettled()**
-   - https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Promise/allSettled
-   - Complete-all behavior, status objects
-
-7. **Node.js: Asynchronous Work**
-   - https://nodejs.dev/en/learn/asynchronous-work/
-   - Async/await patterns, error handling
-
-### Groundswell Codebase
-
-8. **@Task Decorator Implementation**
-   - `/home/dustin/projects/groundswell/src/decorators/task.ts`
-   - Current fail-fast implementation
-
-9. **Error Strategy Types**
-   - `/home/dustin/projects/groundswell/src/types/error-strategy.ts`
-   - Defined but not implemented
-
-10. **Workflow Documentation**
-    - `/home/dustin/projects/groundswell/docs/workflow.md`
-    - Current error handling patterns
-
-11. **Concurrent Tasks Example**
-    - `/home/dustin/projects/groundswell/examples/examples/06-concurrent-tasks.ts`
-    - Real-world usage examples
-
-### Research Context
-
-12. **Error Handling Best Practices Research**
-    - `/home/dustin/projects/groundswell/plan/docs/research/error_handling_patterns.md`
-    - TypeScript error handling patterns, state capture
-
-13. **PRD: Hierarchical Workflow Engine**
-    - `/home/dustin/projects/groundswell/PRPs/001-hierarchical-workflow-engine.md`
-    - Requirements for observability and error handling
-
----
-
-**Document Version**: 1.0
-**Last Updated**: 2026-01-12
-**Status**: Ready for Implementation
-**Next Review**: After Phase 1 Implementation