The ultimate harness for coding agents and beyond. All your agents. All your machines. One command center. Start with code, then follow your curiosity and build across disciplines: CAD, circuits, robots, games and music.
TypeScript 3d-modelingagent-frameworkagent-harnessai-agentscadclaude-codecodexcoding-agentsdeveloper-toolsesp32game-developmentmcp-servermulti-agentopen-hardwareopencodeopenharnesspcbroboticsterminaltmux
import 'dart:async';
import 'dart:convert';
import 'package:flutter/foundation.dart';
import '../core/local_key_value_store.dart';
enum WorkspaceLesson { agent, pane, zoom, commands }
/// Progress belongs to this person's desktop, never to an agent or project.
/// Merely opening a chooser does not complete the agent/pane lessons.
class WorkspaceLearning extends ChangeNotifier {
WorkspaceLearning({this.storage});
final LocalKeyValueStore? storage;
static const storageKey = 'workspace_quick_start_v1';
final _completed = <WorkspaceLesson>{};
bool active = false;
bool dismissed = false;
bool loaded = false;
bool _disposed = false;
int _revision = 0;
Future<void> _saving = Future.value();
Set<WorkspaceLesson> get completed => Set.unmodifiable(_completed);
bool get finished => _completed.length == WorkspaceLesson.values.length;
bool get offer => loaded && !active && !dismissed && !finished;
WorkspaceLesson? get next => WorkspaceLesson.values
.where((step) => !_completed.contains(step))
.firstOrNull;
Future<void> load() async {
final revision = _revision;
try {
final raw = await storage?.read(storageKey);
if (_disposed || revision != _revision) return;
final data = raw == null ? null : jsonDecode(raw);
if (data is Map) {
active = data['active'] == true;
dismissed = data['dismissed'] == true;
final done = data['completed'];
if (done is List) {
_completed.addAll(
WorkspaceLesson.values.where((s) => done.contains(s.name)),
);
}
}
} catch (_) {
// A missing/corrupt preference must never block the workspace.
} finally {
if (!_disposed) {
loaded = true;
notifyListeners();
}
}
}
void start() {
if (finished) _completed.clear();
active = true;
dismissed = false;
_changed();
}
void pause() {
active = false;
dismissed = true;
_changed();
}
void observe({required int agents, required bool zoomed}) {
if (!active || finished) return;
final before = _completed.length;
if (agents > 0) _completed.add(WorkspaceLesson.agent);
if (agents > 1) _completed.add(WorkspaceLesson.pane);
if (agents > 1 && zoomed) _completed.add(WorkspaceLesson.zoom);
if (before != _completed.length) _changed();
}
void commandSearchOpened() {
if (!active || next != WorkspaceLesson.commands) return;
_completed.add(WorkspaceLesson.commands);
_changed();
}
void _changed() {
_revision++;
final data = jsonEncode({
'active': active,
'dismissed': dismissed,
'completed': _completed.map((s) => s.name).toList(),
});
_saving = _saving.then((_) async {
try {
await storage?.write(storageKey, data);
} catch (_) {
// Learning remains available if a preference cannot be saved.
}
});
notifyListeners();
}
Future<void> flush() => _saving;
@override
void dispose() {
_disposed = true;
super.dispose();
}
}