- Fix worker using wrong DB by defaulting to CurrentUser service account
and expanding ~ to absolute paths at install time
- Fix DbContext disposed before fire-and-forget by passing taskId instead
of TaskEntity into RunInSlotAsync, which creates its own context
- Fix ActiveTaskDto property casing mismatch between hub and client
- Move WAL mode PRAGMA before migrations to prevent concurrent lock issues
- Replace FirstAsync with FirstOrDefaultAsync + null guards in tag operations
- Add delete confirmation flow for lists
- Log fire-and-forget exceptions instead of swallowing them
- Broadcast RunCreated event from WorkerHub.RunNow
- Add IDisposable to MainWindowViewModel for event handler cleanup
- Preserve subtask CreatedAt on updates instead of overwriting
- Replace bare catch blocks with Debug.WriteLine logging
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace LiveLines with formatted LiveText + StreamLineFormatter
- Add log reload from disk for completed tasks
- Add start feedback (starting.../running) on detail and list views
- Apply dark theme (WindowBgBrush, AccentBrush) to editor modals
- Add model/system-prompt/agent-path config to ListEditorView
- Wire config loading/saving in MainWindowViewModel
- Fix duplicate AgentInfo DTO (use canonical Data.Models version)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix _startCts lifecycle leak: old CTS is cancelled and disposed before replacement
- Guard StartAsync re-entrancy: early-return if retry loop is still running (lock + IsCompleted check)
- Await _retryLoopTask in both StopAsync and DisposeAsync before stopping/disposing the hub
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add _startCts?.Cancel() at the top of DisposeAsync to prevent
ConnectWithRetryAsync from calling _hub.StartAsync on an already-disposed
HubConnection when the caller disposes without first calling StopAsync.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add IndefiniteRetryPolicy for WithAutomaticReconnect, wrap initial
StartAsync in a retry loop so the UI keeps retrying when Worker is
offline at startup, and expose IsReconnecting to StatusBar
("Connecting..." / "Online" / "Offline").
App: build a ServiceProvider in Program.cs (AppSettings, SqliteConnectionFactory,
all repositories, GitService, WorkerClient, all view-models), apply schema, then
hand control to Avalonia. App.OnFrameworkInitializationCompleted resolves
MainWindowViewModel from the container.
Ui:
- AppSettings POCO loaded from ~/.todo-app/ui.config.json (db path, hub url).
- WorkerClient wraps HubConnection with auto-reconnect, exposes IsConnected and
ActiveTasks plus C# events for TaskStarted/Finished/Message/Updated and
WorktreeUpdated; all inbound events are marshalled to the UI thread.
- ViewModels: MainWindow (lists CRUD via ListEditor dialog), TaskList (load by
list, add/edit/delete, auto WakeQueue on agent+queued create), TaskItem
(RunNow gated on connection + status), TaskDetail (description, result, live
ndjson rolling buffer of 500 lines, worktree branch/diff with merge/keep/
discard via GitService), StatusBar, ListEditor, TaskEditor.
- Views: 3-pane MainWindow (lists | tasks | detail) with GridSplitters, status
bar, dialog windows for the editors. Status badges via StatusColorConverter.
- Markdown rendering, folder picker, delete-confirmation, settings dialog and
scroll-to-bottom on the live log are intentionally TODO -- functional
scaffold only.
Tests: also debounce the FIFO queue test (poll instead of Task.Delay(200)) so
the assertion isn't racy when the suite runs alongside the slower git tests.
38 tests pass.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
