feat(board): master-task board, non-destructive stall watchdog, review flow
CI - Build & Test / Backend (.NET) (push) Successful in 32s
CI - Build & Test / Frontend (Vue/TS) (push) Successful in 18s
CI - Build & Test / Security Check (push) Successful in 2s
CI - Build & Test / Deploy Nexus (push) Has been skipped

Board is now a clean master-task view:
- GetBoardAsync returns only top-level (master) tasks; child-tasks render
  nested inside their parent card instead of as separate column cards, so a
  big task split into many sub-tasks stays one card (orphans treated as master)
- New DoneChildTaskCount on the DTO for real progress bars
- Child/detail consumers (GetChildren endpoint, TaskBridgeService) query
  children directly instead of scraping the flat board

Stall watchdog (replaces destructive auto-reset):
- StaleTaskRecoveryService.FlagStalledInProgressTasksAsync marks In-progress
  tasks with no activity past the threshold as stalled (activity event +
  Iris notification) WITHOUT resetting the column — no work is discarded.
  Idempotent: a task is not re-flagged until real progress happens
- BackgroundService now runs this watchdog (TaskRecovery:StalledMinutes=40,
  interval 10m); hard reset kept only on the explicit manual endpoint

Review flow (Bao/Iris only):
- POST tasks/{id}/approve (Review -> Done)
- POST tasks/{id}/request-changes (Review -> target, mandatory comment,
  ExpectedFrom=iris, notifies Iris)

Frontend:
- BoardCard component: master card with ball chip (who has it), progress from
  children, expand to show children grouped by agent with per-child state +
  stalled marker, stalled chip on the master, review action buttons
- Request-changes modal; tasks store approveReview/requestChanges actions

Tests: watchdog flag/idempotency + review threshold; 135 backend tests green.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
2026-07-10 22:26:50 +02:00
parent f564ecfbc7
commit aef76d5f45
15 changed files with 789 additions and 432 deletions
@@ -2,5 +2,9 @@ namespace Nexus.Api.Services;
public interface IStaleTaskRecoveryService
{
/// <summary>Nicht-destruktiv: markiert hängende In-progress-Tasks und benachrichtigt Iris.</summary>
Task<int> FlagStalledInProgressTasksAsync(TimeSpan stalledThreshold, CancellationToken ct = default);
/// <summary>Destruktiv (nur manuell): setzt hängende In-progress-Tasks hart auf Backlog.</summary>
Task<int> ResetStaleInProgressTasksAsync(TimeSpan staleThreshold, CancellationToken ct = default);
}
+3
View File
@@ -33,9 +33,12 @@ public interface ITaskService
// Task Board
Task<BoardResponse> GetBoardAsync(CancellationToken ct = default);
Task<TaskOperationResult> MoveTaskAsync(Guid id, string newState, CancellationToken ct = default);
Task<TaskOperationResult> ApproveReviewAsync(Guid id, CancellationToken ct = default);
Task<TaskOperationResult> RequestChangesAsync(Guid id, string comment, string? targetState, CancellationToken ct = default);
Task<int> ResetStaleAsync(int staleHours, CancellationToken ct = default);
Task<int> ResetStaleInProgressTasksAsync(TimeSpan staleThreshold, CancellationToken ct = default);
Task<IReadOnlyList<WorkTask>> GetChildTasksAsync(Guid parentId, CancellationToken ct = default);
Task<List<DashboardTaskDto>> GetChildTaskDtosAsync(Guid parentId, CancellationToken ct = default);
Task<List<ActivityEvent>> GetTaskActivityAsync(Guid taskId, CancellationToken ct = default);
Task<DashboardTaskDto?> GetDashboardTaskByIdAsync(Guid id, CancellationToken ct = default);
@@ -13,9 +13,9 @@ public sealed class StaleTaskRecoveryBackgroundService(
{
try
{
var resetCount = await RunRecoveryOnceAsync(stoppingToken);
if (resetCount > 0)
logger.LogInformation("Stale task recovery reset {ResetCount} task(s).", resetCount);
var flaggedCount = await RunWatchdogOnceAsync(stoppingToken);
if (flaggedCount > 0)
logger.LogInformation("Stall watchdog flagged {FlaggedCount} stalled task(s) for Iris.", flaggedCount);
}
catch (OperationCanceledException) when (stoppingToken.IsCancellationRequested)
{
@@ -37,10 +37,10 @@ public sealed class StaleTaskRecoveryBackgroundService(
}
}
public async Task<int> RunRecoveryOnceAsync(CancellationToken ct = default)
public async Task<int> RunWatchdogOnceAsync(CancellationToken ct = default)
{
await using var scope = scopeFactory.CreateAsyncScope();
var recoveryService = scope.ServiceProvider.GetRequiredService<IStaleTaskRecoveryService>();
return await recoveryService.ResetStaleInProgressTasksAsync(optionsMonitor.CurrentValue.GetStaleThreshold(), ct);
return await recoveryService.FlagStalledInProgressTasksAsync(optionsMonitor.CurrentValue.GetStalledThreshold(), ct);
}
}
+9 -1
View File
@@ -4,8 +4,16 @@ public sealed class StaleTaskRecoveryOptions
{
public const string SectionName = "TaskRecovery";
/// <summary>Schwelle (Minuten) ohne Aktivität, ab der ein In-progress-Task als hängend gilt.</summary>
public int StalledMinutes { get; set; } = 40;
/// <summary>Prüfintervall des Watchdogs.</summary>
public int IntervalMinutes { get; set; } = 10;
/// <summary>Nur für den manuellen Hard-Reset-Endpoint: Alter (Stunden) ab dem hart zurückgesetzt wird.</summary>
public int StaleHours { get; set; } = 2;
public int IntervalMinutes { get; set; } = 30;
public TimeSpan GetStalledThreshold() => TimeSpan.FromMinutes(Math.Max(1, StalledMinutes));
public TimeSpan GetStaleThreshold() => TimeSpan.FromHours(Math.Max(1, StaleHours));
+79 -97
View File
@@ -7,8 +7,81 @@ namespace Nexus.Api.Services;
public sealed class StaleTaskRecoveryService(
ITaskRepository taskRepository,
IActivityRepository activityRepository,
ILiveUpdateService liveUpdateService) : IStaleTaskRecoveryService
ILiveUpdateService liveUpdateService,
INotificationService notificationService) : IStaleTaskRecoveryService
{
private const string StalledActivityType = "stalled";
/// <summary>
/// NICHT-destruktiver Watchdog: markiert „In progress"-Tasks ohne Aktivität seit
/// <paramref name="stalledThreshold"/> als hängend (Activity-Event + Notification an Iris),
/// OHNE die Spalte zu ändern oder Arbeit zu verwerfen. Iris eskaliert dann (nachfragen,
/// neu delegieren, ggf. auf Blocked setzen). Dedup: bereits gemeldete Hänger werden nicht
/// erneut gemeldet, solange kein neuer Fortschritt (andere Activity) dazwischen liegt.
/// </summary>
public async Task<int> FlagStalledInProgressTasksAsync(TimeSpan stalledThreshold, CancellationToken ct = default)
{
var now = DateTimeOffset.UtcNow;
var threshold = now - stalledThreshold;
var allTasks = await taskRepository.GetAllAsync(ct);
var inProgress = allTasks
.Where(t => string.Equals(t.State, TaskStateHelper.ToStateString(TaskState.InProgress), StringComparison.OrdinalIgnoreCase))
.ToList();
if (inProgress.Count == 0)
return 0;
var activities = await activityRepository.GetRecentForTasksAsync(inProgress.Select(t => t.Id), ct);
var activityByTask = activities
.Where(a => a.TaskId.HasValue)
.GroupBy(a => a.TaskId!.Value)
.ToDictionary(g => g.Key, g => g.OrderByDescending(a => a.CreatedAt).ToList());
var flaggedCount = 0;
foreach (var task in inProgress)
{
activityByTask.TryGetValue(task.Id, out var taskActivity);
var latest = taskActivity?.FirstOrDefault();
var lastProgressAt = latest?.CreatedAt ?? task.UpdatedAt;
if (lastProgressAt >= threshold)
continue;
// Dedup: schon als hängend gemeldet und seither kein neuer Fortschritt.
if (latest is not null && string.Equals(latest.Type, StalledActivityType, StringComparison.OrdinalIgnoreCase))
continue;
var silentFor = now - lastProgressAt;
await activityRepository.AddAsync(new ActivityEvent
{
Type = StalledActivityType,
Message = $"Watchdog: keine Aktivität seit {FormatDuration(silentFor)} (Schwelle {FormatDuration(stalledThreshold)}). Task bleibt In progress, Iris zur Eskalation benachrichtigt.",
TaskId = task.Id
}, ct);
await notificationService.CreateAsync(
"task_stalled",
$"Task hängt: {task.Title}",
$"Seit {FormatDuration(silentFor)} keine Aktivität. Bitte nachfassen, neu delegieren oder blockieren.",
"iris",
task.Id,
ct);
flaggedCount++;
}
if (flaggedCount > 0)
liveUpdateService.Publish("tasks.board.snapshot", await BuildBoardSnapshotAsync(ct), "board");
return flaggedCount;
}
/// <summary>
/// Destruktiver Fallback (nur manuell via Endpoint / expliziter Cron): setzt hängende
/// „In progress"-Tasks hart auf Backlog zurück. Verwirft laufenden Kontext — daher NICHT
/// mehr der Standard-Watchdog, sondern nur noch auf Anforderung.
/// </summary>
public async Task<int> ResetStaleInProgressTasksAsync(TimeSpan staleThreshold, CancellationToken ct = default)
{
var threshold = DateTimeOffset.UtcNow - staleThreshold;
@@ -80,88 +153,8 @@ public sealed class StaleTaskRecoveryService(
private async Task<BoardResponse> BuildBoardSnapshotAsync(CancellationToken ct)
{
var allTasks = await taskRepository.GetAllAsync(ct);
var taskIds = allTasks.Select(task => task.Id).ToList();
var activity = await activityRepository.GetRecentForTasksAsync(taskIds, ct);
var backlog = new List<DashboardTaskDto>();
var inProgress = new List<DashboardTaskDto>();
var review = new List<DashboardTaskDto>();
var blocked = new List<DashboardTaskDto>();
var done = new List<DashboardTaskDto>();
foreach (var task in allTasks)
{
var dto = MapToDtoWithChildren(task, allTasks, activity);
switch (task.State.ToLowerInvariant())
{
case "backlog": backlog.Add(dto); break;
case "in progress": inProgress.Add(dto); break;
case "review": review.Add(dto); break;
case "blocked": blocked.Add(dto); break;
case "done": done.Add(dto); break;
default: backlog.Add(dto); break;
}
}
backlog.Sort(SortByPriorityThenCreatedAt);
inProgress.Sort(SortByPriorityThenCreatedAt);
review.Sort(SortByPriorityThenCreatedAt);
blocked.Sort(SortByPriorityThenCreatedAt);
done.Sort(SortByPriorityThenCreatedAt);
return new BoardResponse(backlog, inProgress, review, blocked, done);
}
private static DashboardTaskDto MapToDtoWithChildren(
WorkTask task,
IReadOnlyList<WorkTask> allTasks,
IEnumerable<ActivityEvent> activity)
{
var childTasks = allTasks
.Where(candidate => candidate.ParentTaskId == task.Id)
.OrderByDescending(candidate => candidate.UpdatedAt)
.ToList();
var childDtos = childTasks.Select(child => MapToDtoWithActivity(child, activity)).ToList();
var openChildTaskCount = childTasks.Count(child => !string.Equals(child.State, "Done", StringComparison.OrdinalIgnoreCase));
var dto = MapToDtoWithActivity(task, activity);
return dto with
{
ChildTasks = childDtos,
ChildTaskCount = childDtos.Count,
OpenChildTaskCount = openChildTaskCount,
HasVisibleDelegation = dto.ParentTaskId.HasValue || childDtos.Count > 0 || dto.IsAgentTask
};
}
private static DashboardTaskDto MapToDtoWithActivity(WorkTask task, IEnumerable<ActivityEvent> activity)
{
var last = activity
.Where(entry => entry.TaskId == task.Id)
.OrderByDescending(entry => entry.CreatedAt)
.FirstOrDefault();
return new DashboardTaskDto(
task.Id,
task.Title,
task.Detail,
task.Source,
task.State,
task.Priority,
task.AssignedTo,
task.ParentTaskId,
task.DueDate,
task.CreatedAt,
task.UpdatedAt,
task.IsAgentTask,
task.ExpectedFrom,
last?.Message,
last?.CreatedAt,
null,
0,
0,
task.ParentTaskId.HasValue || task.IsAgentTask);
var activity = await activityRepository.GetRecentForTasksAsync(allTasks.Select(task => task.Id), ct);
return TaskService.BuildMasterBoard(allTasks, activity);
}
private static string BuildActivityMessage(
@@ -190,20 +183,9 @@ public sealed class StaleTaskRecoveryService(
}
private static string FormatDuration(TimeSpan duration)
=> duration.ToString(@"dd\.hh\:mm\:ss");
private static int SortByPriorityThenCreatedAt(DashboardTaskDto a, DashboardTaskDto b)
{
var priorityCompare = PriorityScore(b.Priority).CompareTo(PriorityScore(a.Priority));
return priorityCompare != 0 ? priorityCompare : a.CreatedAt.CompareTo(b.CreatedAt);
if (duration.TotalHours >= 1)
return $"{(int)duration.TotalHours}h {duration.Minutes}min";
return $"{Math.Max(0, (int)duration.TotalMinutes)}min";
}
private static int PriorityScore(string priority) => priority.ToLowerInvariant() switch
{
"high" => 3,
"medium" => 2,
"normal" => 2,
"low" => 1,
_ => 2
};
}
+1 -14
View File
@@ -214,13 +214,7 @@ public sealed class TaskBridgeService(
public async Task<IReadOnlyList<DashboardTaskDto>> GetChildTasksAsync(
Guid parentTaskId, CancellationToken ct = default)
{
var board = await taskService.GetBoardAsync(ct);
return FlattenBoard(board)
.Where(task => task.ParentTaskId == parentTaskId)
.OrderByDescending(task => task.UpdatedAt)
.ToList();
}
=> await taskService.GetChildTaskDtosAsync(parentTaskId, ct);
public async Task<List<ActivityEvent>> GetTaskActivityAsync(
Guid taskId, CancellationToken ct = default)
@@ -250,13 +244,6 @@ public sealed class TaskBridgeService(
return AgentIdentityCatalog.NormalizeActorId(actorId, allowedActors);
}
private static IEnumerable<DashboardTaskDto> FlattenBoard(BoardResponse board)
=> board.Offen
.Concat(board.InProgress)
.Concat(board.Review)
.Concat(board.Blocked)
.Concat(board.Done);
private static DashboardTaskDto MapToDto(WorkTask t) => new(
t.Id, t.Title, t.Detail, t.Source, t.State, t.Priority, t.AssignedTo,
t.ParentTaskId, t.DueDate, t.CreatedAt, t.UpdatedAt,
+106 -6
View File
@@ -422,6 +422,19 @@ public sealed class TaskService(
{
var all = (await taskRepo.GetAllAsync(ct)).ToList();
var activity = await activityRepo.GetRecentForTasksAsync(all.Select(t => t.Id), ct);
return BuildMasterBoard(all, activity);
}
/// <summary>
/// Baut das Board aus NUR den Master-Tasks (Top-Level). Child-Tasks erscheinen
/// nicht als eigene Karten, sondern verschachtelt in ihrem Parent — so bleibt das
/// Board übersichtlich, auch wenn Iris eine große Aufgabe in viele Teilaufgaben
/// zerlegt. Waisen (Parent existiert nicht mehr) werden als Master behandelt,
/// damit nichts unsichtbar wird.
/// </summary>
internal static BoardResponse BuildMasterBoard(IReadOnlyList<WorkTask> all, IReadOnlyList<ActivityEvent> activity)
{
var ids = all.Select(t => t.Id).ToHashSet();
var offen = new List<DashboardTaskDto>();
var inProgress = new List<DashboardTaskDto>();
@@ -431,9 +444,10 @@ public sealed class TaskService(
foreach (var task in all)
{
// Ohne verschachtelte Child-DTOs: Children sind als eigene Karten im Board,
// die Nested-Duplikate haben die Payload nur verdoppelt (Counts bleiben).
var dto = MapToDtoWithChildren(task, all, activity, includeChildren: false);
var isMaster = !task.ParentTaskId.HasValue || !ids.Contains(task.ParentTaskId.Value);
if (!isMaster) continue;
var dto = MapToDtoWithChildren(task, all, activity, includeChildren: true);
switch (task.State.ToLowerInvariant())
{
case "backlog": offen.Add(dto); break;
@@ -492,6 +506,78 @@ public sealed class TaskService(
return await UpdateTaskStatusInternalAsync(task, canonical, caller, "task", $"Task \"{task.Title}\" moved to {canonical}", ct);
}
/// <summary>
/// Review-Abnahme durch Bao/Iris: Review → Done. Nur aus dem Review-Status erlaubt.
/// </summary>
public async Task<TaskOperationResult> ApproveReviewAsync(Guid id, CancellationToken ct = default)
{
var task = await taskRepo.GetByIdAsync(id, ct);
if (task is null) return new TaskOperationResult(TaskOperationOutcome.NotFound);
var caller = ResolveCaller();
if (!TaskStateHelper.CanChangeState(caller, task))
return new TaskOperationResult(TaskOperationOutcome.InvalidState);
if (!string.Equals(task.State, "Review", StringComparison.OrdinalIgnoreCase))
return new TaskOperationResult(TaskOperationOutcome.InvalidState, task);
task.ExpectedFrom = null;
return await UpdateTaskStatusInternalAsync(
task,
TaskStateHelper.ToStateString(TaskState.Done),
caller,
"review",
$"Review abgenommen von {caller}: \"{task.Title}\" → Done",
ct);
}
/// <summary>
/// Änderung anfordern: Review → Zielspalte (Default In progress) mit Pflichtkommentar.
/// Setzt ExpectedFrom=iris und benachrichtigt sie, damit sie autonom nacharbeitet.
/// </summary>
public async Task<TaskOperationResult> RequestChangesAsync(Guid id, string comment, string? targetState, CancellationToken ct = default)
{
var task = await taskRepo.GetByIdAsync(id, ct);
if (task is null) return new TaskOperationResult(TaskOperationOutcome.NotFound);
var caller = ResolveCaller();
if (!TaskStateHelper.CanChangeState(caller, task))
return new TaskOperationResult(TaskOperationOutcome.InvalidState);
if (!string.Equals(task.State, "Review", StringComparison.OrdinalIgnoreCase))
return new TaskOperationResult(TaskOperationOutcome.InvalidState, task);
var target = TaskStateHelper.AllStates.FirstOrDefault(s => s.Equals(targetState, StringComparison.OrdinalIgnoreCase))
?? TaskStateHelper.ToStateString(TaskState.InProgress);
// Aus dem Review geht es zurück in die Arbeit — nie direkt nach Done oder Review.
if (string.Equals(target, "Done", StringComparison.OrdinalIgnoreCase)
|| string.Equals(target, "Review", StringComparison.OrdinalIgnoreCase))
target = TaskStateHelper.ToStateString(TaskState.InProgress);
var trimmed = comment.Trim();
await activityRepo.AddAsync(new ActivityEvent
{
Type = "review_changes_requested",
Message = $"Änderung angefordert von {caller}: {trimmed}",
TaskId = task.Id
}, ct);
task.ExpectedFrom = "iris";
var result = await UpdateTaskStatusInternalAsync(
task, target, caller, "review",
$"Review zurückgegeben von {caller} → {target}", ct);
await notificationService.CreateAsync(
"task_changes_requested",
$"Änderung angefordert: {task.Title}",
trimmed,
"iris",
task.Id,
ct);
return result;
}
public Task<int> ResetStaleAsync(int staleHours, CancellationToken ct = default)
{
var normalizedHours = Math.Max(1, staleHours);
@@ -509,20 +595,33 @@ public sealed class TaskService(
.ToList();
}
/// <summary>
/// Child-Tasks eines Parents als DTOs — direkt aus dem Repo, nicht aus dem Board
/// (das zeigt Children ja nur noch verschachtelt an). Für Detailansicht + Bridge.
/// </summary>
public async Task<List<DashboardTaskDto>> GetChildTaskDtosAsync(Guid parentId, CancellationToken ct = default)
{
var all = (await taskRepo.GetAllAsync(ct)).ToList();
var activity = await activityRepo.GetRecentForTasksAsync(all.Select(t => t.Id), ct);
return all.Where(t => t.ParentTaskId == parentId)
.OrderByDescending(t => t.UpdatedAt)
.Select(child => MapToDtoWithChildren(child, all, activity, includeChildren: false))
.ToList();
}
public async Task<List<ActivityEvent>> GetTaskActivityAsync(Guid taskId, CancellationToken ct = default)
{
var all = await activityRepo.GetRecentAsync(100, ct);
return all.Where(e => e.TaskId == taskId).ToList();
}
private DashboardTaskDto MapToDtoWithChildren(WorkTask task, IReadOnlyList<WorkTask> allTasks, IEnumerable<ActivityEvent> activity, bool includeChildren = true)
private static DashboardTaskDto MapToDtoWithChildren(WorkTask task, IReadOnlyList<WorkTask> allTasks, IEnumerable<ActivityEvent> activity, bool includeChildren = true)
{
var childTasks = allTasks.Where(t => t.ParentTaskId == task.Id)
.OrderByDescending(t => t.UpdatedAt)
.ToList();
// includeChildren=false (Board/SSE-Snapshot): Children erscheinen dort ohnehin
// als eigene Karten — verschachtelte Child-DTOs verdoppeln nur die Payload.
// includeChildren=false: nur Zähler, keine verschachtelten Child-DTOs (schlanke Payload).
var childDtos = includeChildren
? childTasks.Select(child => MapToDtoWithActivity(child, activity, allTasks)).ToList()
: null;
@@ -534,6 +633,7 @@ public sealed class TaskService(
ChildTasks = childDtos,
ChildTaskCount = childTasks.Count,
OpenChildTaskCount = openChildTaskCount,
DoneChildTaskCount = childTasks.Count - openChildTaskCount,
HasVisibleDelegation = dto.ParentTaskId.HasValue || childTasks.Count > 0 || dto.IsAgentTask
};
}