Files
clide/test/builtin/claude/claude_pane_test.dart
T
jpmschweitzerandClaude Opus 4.8 b399acedbf surface Claude Code Workflow runs in convo + sidebar (T-416)
A `Workflow` tool-use launches its multi-agent run in the background and
returns immediately; the real fan-out arrives out-of-band on stream-json
`type:"system"` task_* events (task_started / task_progress / task_updated /
task_notification) keyed by the launching tool-use id — which clide was
dropping. (Wire shape captured by two live stream-json probes; recorded on
the ticket.)

- workflow_run.dart: a pure, Flutter-free WorkflowRun/WorkflowAgent model
  that folds those events (phases, per-agent start→progress→done deltas,
  usage) into a snapshot.
- StreamJsonSession recognises the events, accumulates a
  Map<toolUseId, WorkflowRun>, and exposes `workflows` + `workflowsStream`.
- A `Workflow` tool-use with a live run renders a dedicated run card —
  phase groups, per-agent rows with spinner/check status, usage, and the
  script — falling back to the generic tool card pre-progress or on reload.
  The run breaks the activity cluster so it's always first-class (like T-342).
- The sidebar Activity tab adds a WORKFLOWS section: one row per run with its
  done/total agent count, tinted by running/done state.

Closes T-416 and the T-410 epic (all children done).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-12 21:44:41 +02:00

470 lines
19 KiB
Dart

/// Widget tests for [ClaudePane] — the spawn/rebind lifecycle (T-269) plus the
/// composer-driven command handling (/clear, /fork, send, mode cycle) and the
/// status / prompt render paths.
///
/// Harness: a fake [ClaudeSessionOrchestrator] (no real `claude` process) and a
/// connected fake IPC that answers `files.root`. The pane awaits a real
/// File(...).exists() transcript probe during spawn, so the spawn/respawn
/// phases run inside tester.runAsync (fake-async would trap that I/O).
library;
import 'dart:async';
import 'dart:convert';
import 'package:clide/builtin/claude/src/claude_composer.dart';
import 'package:clide/builtin/claude/src/claude_pane.dart';
import 'package:clide/builtin/claude/src/conversation_view.dart';
import 'package:clide/builtin/claude/src/model_picker_card.dart';
import 'package:clide/builtin/claude/src/session_naming.dart';
import 'package:clide/builtin/claude/src/session_orchestrator.dart';
import 'package:clide/builtin/claude/src/session_picker.dart';
import 'package:clide/builtin/claude/src/stream_json_session.dart';
import 'package:clide/clide.dart';
import 'package:clide/kernel/kernel.dart';
import 'package:flutter/services.dart';
import 'package:flutter/widgets.dart';
import 'package:flutter_test/flutter_test.dart';
import '../../helpers/kernel_fixture.dart';
class _FakeProc extends StreamJsonProcess {
final _ctl = StreamController<String>.broadcast();
final List<String> writes = [];
bool killed = false;
@override
Stream<String> get lines => _ctl.stream;
@override
void writeLine(String line) => writes.add(line);
@override
Future<void> kill() async {
killed = true;
if (!_ctl.isClosed) await _ctl.close();
}
void feed(Map<String, Object?> event) {
if (!_ctl.isClosed) _ctl.add(jsonEncode(event));
}
}
void main() {
late KernelFixture f;
late ClaudeSessionOrchestrator orch;
late String root;
final created = <_FakeProc>[];
final spawnArgs = <List<String>>[];
setUp(() async {
f = await KernelFixture.create();
created.clear();
spawnArgs.clear();
root = '/repo-a';
orch = ClaudeSessionOrchestrator(
processFactory: ({required sessionArgs, required cwd, env}) async {
final p = _FakeProc();
created.add(p);
spawnArgs.add(sessionArgs);
return p;
},
);
activeSessionOrchestrator = orch;
f.ipc.setConnected(true);
f.ipc.stub('files.root', (_) async => IpcResponse.ok(id: '', data: {'path': root}));
});
tearDown(() async {
activeSessionOrchestrator = null;
orch.dispose();
await f.dispose();
});
Widget tree(ClaudePane pane) => Directionality(
textDirection: TextDirection.ltr,
child: ClideKernel(
services: f.services,
child: ClideTheme(
controller: f.services.theme,
child: MediaQuery(
data: const MediaQueryData(),
child: Align(
alignment: Alignment.topLeft,
child: SizedBox(
width: 900,
height: 700,
child: DialogHost(
router: f.services.dialog,
child: Overlay(initialEntries: [OverlayEntry(builder: (_) => pane)]),
),
),
),
),
),
),
);
// Pump the pane and release its project-wait gate so _spawn runs. The whole
// chain (incl. the real transcript-probe I/O) runs in the real zone.
Future<void> mount(WidgetTester tester, ClaudePane pane, {String? openPath}) async {
await tester.runAsync(() async {
await tester.pumpWidget(tree(pane));
f.services.events.emit(ProjectOpened(path: openPath ?? root));
await Future<void>.delayed(const Duration(milliseconds: 60));
});
await tester.pump();
}
// Run [fn] (a stream feed or an async command) in the real zone, then settle.
Future<void> act(WidgetTester tester, FutureOr<void> Function() fn) async {
await tester.runAsync(() async {
await fn();
await Future<void>.delayed(const Duration(milliseconds: 60));
});
await tester.pump();
}
ClaudeComposer composer(WidgetTester tester) => tester.widget<ClaudeComposer>(find.byType(ClaudeComposer));
testWidgets('spawns a session bound to the active repo and renders', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final primary = orch.byId('primary');
expect(primary, isNotNull);
expect(primary!.cwd, '/repo-a');
expect(primary.sessionId, primarySessionId('/repo-a'));
expect(find.byType(ConversationView), findsOneWidget);
expect(find.byType(ClaudeComposer), findsOneWidget);
});
testWidgets('switching the workspace in place rebinds to the new repo (T-269)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
expect(orch.byId('primary')!.cwd, '/repo-a');
final firstProc = created.single;
root = '/repo-b';
await act(tester, () => f.services.events.emit(const ProjectOpened(path: '/repo-b')));
final primary = orch.byId('primary')!;
expect(primary.cwd, '/repo-b');
expect(primary.sessionId, primarySessionId('/repo-b'));
expect(firstProc.killed, isTrue);
expect(created, hasLength(2));
});
testWidgets('a re-open of the same repo does not respawn', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
expect(created, hasLength(1));
await act(tester, () => f.services.events.emit(ProjectOpened(path: root)));
expect(created, hasLength(1), reason: 'same path → no rebind');
});
testWidgets('sending a message writes to the session', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
composer(tester).onSubmit('hello there');
await tester.pump();
expect(proc.writes.any((w) => w.contains('hello there')), isTrue);
});
testWidgets('a Workflow run renders its dedicated card in the conversation (T-416)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
final semantics = tester.ensureSemantics();
await act(tester, () {
proc.feed({
'type': 'assistant',
'uuid': 'a-wf',
'message': {
'role': 'assistant',
'content': [
{
'type': 'tool_use',
'id': 'toolu_wf',
'name': 'Workflow',
'input': {'script': 'await parallel([])'},
},
],
},
});
proc.feed({
'type': 'system',
'subtype': 'task_progress',
'tool_use_id': 'toolu_wf',
'workflow_progress': [
{'type': 'workflow_agent', 'index': 1, 'label': 'first agent', 'state': 'done'},
{'type': 'workflow_agent', 'index': 2, 'label': 'second agent', 'state': 'start'},
],
});
});
// The dedicated workflow collapser, with its live done/total agent counter.
expect(find.bySemanticsLabel('workflow, 1/2 agents, collapsed'), findsOneWidget);
semantics.dispose();
});
testWidgets('/clear empties the deterministic session in place', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final firstProc = created.single;
final id = orch.byId('primary')!.sessionId;
await act(tester, () => composer(tester).onSubmit('/clear'));
expect(firstProc.killed, isTrue);
expect(created, hasLength(2));
// Re-bound to the SAME deterministic id (cleared in place, not a random id).
expect(orch.byId('primary')!.sessionId, id);
});
testWidgets('/clear in a fork pane clears instead of re-forking (T-375)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false, isPrimary: false, secondaryIndex: 1, forkSourceId: 'source-session-uuid'));
// First bind forks from the source.
expect(spawnArgs.single, containsAll(['--fork-session', 'source-session-uuid']));
await act(tester, () => composer(tester).onSubmit('/clear'));
// The respawn must NOT fork the original again — the fork source is a
// one-shot spawn parameter consumed by the first bind.
expect(spawnArgs, hasLength(2));
expect(spawnArgs.last, isNot(contains('--fork-session')));
expect(spawnArgs.last, isNot(contains('source-session-uuid')));
});
testWidgets('/fork delegates to the onFork callback with the session id', (tester) async {
String? forkedWith;
await mount(tester, ClaudePane(showChrome: false, onFork: (sid) => forkedWith = sid));
composer(tester).onSubmit('/fork');
await tester.pump();
expect(forkedWith, primarySessionId('/repo-a'));
});
testWidgets('/model with an argument sends set_model, never a user message (T-408)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
await act(tester, () => composer(tester).onSubmit('/model sonnet'));
final sent = proc.writes.map((w) => jsonDecode(w) as Map).toList();
final setModel = sent.where((m) => (m['request'] as Map?)?['subtype'] == 'set_model').toList();
expect(setModel, hasLength(1));
expect((setModel.single['request'] as Map)['model'], 'sonnet');
expect(sent.any((m) => m['type'] == 'user'), isFalse, reason: '/model must not be forwarded as message text');
expect(find.byType(ModelPickerCard), findsNothing);
});
testWidgets('bare /model swaps the picker in; picking sends set_model and restores the composer (T-408)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
await act(tester, () => composer(tester).onSubmit('/model'));
expect(find.byType(ModelPickerCard), findsOneWidget);
expect(find.byType(ClaudeComposer), findsNothing, reason: 'the picker takes the interaction zone (D-78)');
tester.widget<ModelPickerCard>(find.byType(ModelPickerCard)).onPick('opus');
await tester.pump();
expect(find.byType(ModelPickerCard), findsNothing);
expect(find.byType(ClaudeComposer), findsOneWidget);
final setModel = proc.writes.map((w) => jsonDecode(w) as Map).where((m) => (m['request'] as Map?)?['subtype'] == 'set_model').toList();
expect((setModel.single['request'] as Map)['model'], 'opus');
});
testWidgets('Esc cancels the /model picker without sending (T-408)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
await act(tester, () => composer(tester).onSubmit('/model'));
expect(find.byType(ModelPickerCard), findsOneWidget);
await tester.sendKeyEvent(LogicalKeyboardKey.escape);
await tester.pump();
expect(find.byType(ModelPickerCard), findsNothing);
expect(find.byType(ClaudeComposer), findsOneWidget);
expect(proc.writes.any((w) => w.contains('set_model')), isFalse);
});
testWidgets('cycling permission mode sends a control message', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
composer(tester).onCycleMode!();
await tester.pump();
expect(proc.writes.any((w) => w.contains('permission')), isTrue);
});
testWidgets('composer draft is retained then cleared', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final c = composer(tester);
c.onDraftChanged!(const TextEditingValue(text: 'a draft'));
c.onDraftChanged!(TextEditingValue.empty);
await tester.pump();
// No throw / no crash; draft round-trips through the pane's per-session map.
expect(find.byType(ClaudeComposer), findsOneWidget);
});
testWidgets('an init event populates the status line', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
await act(
tester,
() => proc.feed({'type': 'system', 'subtype': 'init', 'model': 'claude-opus-4-8', 'permissionMode': 'plan', 'session_id': primarySessionId('/repo-a')}),
);
// The init event flows through the session into the pane's status path
// (the rendered slot lives in the status bar, absent from this harness).
expect(orch.byId('primary')!.session.status.permissionMode, 'plan');
expect(orch.byId('primary')!.session.status.model, 'claude-opus-4-8');
});
testWidgets('a can_use_tool request renders a prompt card in place of the composer', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
await act(
tester,
() => proc.feed({
'type': 'control_request',
'request_id': 'req-1',
'request': {
'subtype': 'can_use_tool',
'tool_name': 'Bash',
'tool_use_id': 'tu-1',
'input': {'command': 'ls'},
},
}),
);
expect(find.byType(ClaudeComposer), findsNothing, reason: 'prompt takes the composer slot (D-78)');
});
testWidgets('a disconnected daemon surfaces an error instead of spawning', (tester) async {
f.ipc.setConnected(false);
await mount(tester, const ClaudePane(showChrome: false));
expect(orch.byId('primary'), isNull);
expect(find.textContaining('not connected'), findsOneWidget);
});
testWidgets('a secondary pane spawns a fresh session under its own key', (tester) async {
await mount(tester, const ClaudePane(isPrimary: false, secondaryIndex: 1, showChrome: false));
final secondary = orch.byId('secondary-1');
expect(secondary, isNotNull);
expect(secondary!.cwd, '/repo-a');
// Secondaries get a fresh random id, not the deterministic primary one.
expect(secondary.sessionId, isNot(primarySessionId('/repo-a')));
// Disposing the secondary closes its session (the !isPrimary branch).
// Pump a different widget type so the pane's State is torn down, not reused.
await tester.pumpWidget(const SizedBox());
await tester.pump();
expect(orch.byId('secondary-1'), isNull);
});
testWidgets('tapping the conversation area focuses the composer', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
await tester.tap(find.byType(ConversationView));
await tester.pump();
expect(find.byType(ClaudeComposer), findsOneWidget);
});
testWidgets('/resume opens the session picker; cancelling leaves the session', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final proc = created.single;
await act(tester, () => composer(tester).onSubmit('/resume'));
expect(find.byType(SessionPickerDialog), findsOneWidget);
// Cancel the picker → no rebind, original session untouched.
await act(tester, () => f.services.dialog.dismiss());
expect(find.byType(SessionPickerDialog), findsNothing);
expect(created, hasLength(1));
expect(proc.killed, isFalse);
});
// ---- T-410 epic: bus-driven owned commands (T-412/T-413/T-414) ----------
// The sidebar controls publish slash-command text on builtin.claude/command;
// the primary pane executes it through _send — the same path as typing.
group('command bus → owned slash commands', () {
Future<void> command(WidgetTester tester, String text) => act(tester, () => f.services.messages.publish('builtin.claude', 'command', {'text': text}));
testWidgets('/effort <level> respawns the session carrying --effort (T-412)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
expect(created, hasLength(1));
await command(tester, '/effort xhigh');
expect(created, hasLength(2), reason: 'effort change = respawn');
final args = spawnArgs.last;
final i = args.indexOf('--effort');
expect(i, isNonNegative, reason: 'args: $args');
expect(args[i + 1], 'xhigh');
// The pane records the level on the session status (the wire never
// reports effort).
expect(orch.byId('primary')!.session.status.effort, 'xhigh');
});
testWidgets('/effort with an unknown level notices and does NOT respawn (T-412)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
await command(tester, '/effort warp9');
expect(created, hasLength(1));
expect(find.textContaining('unknown effort "warp9"'), findsOneWidget);
});
testWidgets('bare /effort opens the effort picker in the interaction zone (T-412)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
await command(tester, '/effort');
expect(find.byType(ModelPickerCard), findsOneWidget);
expect(find.text('effort'), findsOneWidget); // the picker title
});
testWidgets('/permissions <mode> sends set_permission_mode (T-413)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
await command(tester, '/permissions plan');
final proc = created.single;
expect(proc.writes.any((w) => w.contains('set_permission_mode') && w.contains('"plan"')), isTrue, reason: proc.writes.join('\n'));
});
testWidgets('bare /permissions opens the mode picker (T-413)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
await command(tester, '/permissions');
expect(find.byType(ModelPickerCard), findsOneWidget);
expect(find.text('permissions'), findsOneWidget);
});
testWidgets('/status and /config navigate the Claude sidebar (T-413)', (tester) async {
final tabs = <String?>[];
final sub = f.services.messages.subscribe(publisher: 'builtin.claude', channel: 'meta.tab').listen((m) => tabs.add(m.data['tab'] as String?));
addTearDown(sub.cancel);
await mount(tester, const ClaudePane(showChrome: false));
await command(tester, '/status');
await command(tester, '/config');
await command(tester, '/mcp');
expect(tabs, ['activity', 'config', 'config']);
});
testWidgets('/memory opens CLAUDE.md in the editor (T-413)', (tester) async {
final opened = <String?>[];
f.ipc.stub('editor.open', (args) async {
opened.add(args['path'] as String?);
return IpcResponse.ok(id: '', data: const {});
});
await mount(tester, const ClaudePane(showChrome: false));
await command(tester, '/memory');
expect(opened, ['/repo-a/CLAUDE.md']);
});
testWidgets('/help renders a local clide summary card (T-413)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
await command(tester, '/help');
expect(find.textContaining('clide commands:'), findsOneWidget);
expect(find.text('clide'), findsOneWidget); // synthetic card attribution
});
testWidgets('a TUI-only command becomes a notice card, never reaching the session (T-411)', (tester) async {
await mount(tester, const ClaudePane(showChrome: false));
final before = created.single.writes.length;
await command(tester, '/doctor');
expect(find.textContaining('/doctor is a Claude Code TUI command'), findsOneWidget);
expect(created.single.writes.length, before, reason: 'nothing forwarded');
});
});
}