test: expand claude work sync live scenarios
This commit is contained in:
parent
7cb6cddee8
commit
bfd8b30ad5
1 changed files with 291 additions and 204 deletions
|
|
@ -49,6 +49,26 @@ const DEFAULT_ORCHESTRATOR_CLI = '/Users/belief/dev/projects/claude/agent_teams_
|
||||||
const DEFAULT_MODEL = 'sonnet';
|
const DEFAULT_MODEL = 'sonnet';
|
||||||
const DEFAULT_EFFORT = 'low' as const;
|
const DEFAULT_EFFORT = 'low' as const;
|
||||||
|
|
||||||
|
type ClaudeStopHookLiveScenarioState = 'still_working' | 'caught_up';
|
||||||
|
|
||||||
|
interface ClaudeStopHookLiveScenarioContext {
|
||||||
|
marker: string;
|
||||||
|
memberName: string;
|
||||||
|
teamName: string;
|
||||||
|
controlUrl: string;
|
||||||
|
taskId?: string;
|
||||||
|
}
|
||||||
|
|
||||||
|
interface ClaudeStopHookLiveScenario {
|
||||||
|
markerSuffix: string;
|
||||||
|
subjectPrefix: string;
|
||||||
|
expectedState: ClaudeStopHookLiveScenarioState;
|
||||||
|
expectedTaskStatus: 'in_progress' | 'completed';
|
||||||
|
expectedMarkerText(marker: string): string;
|
||||||
|
buildTaskPromptLines(context: ClaudeStopHookLiveScenarioContext): string[];
|
||||||
|
buildInstructionLines(context: Required<ClaudeStopHookLiveScenarioContext>): string[];
|
||||||
|
}
|
||||||
|
|
||||||
liveDescribe('Member work sync Claude Stop hook live e2e', () => {
|
liveDescribe('Member work sync Claude Stop hook live e2e', () => {
|
||||||
let tempDir: string;
|
let tempDir: string;
|
||||||
let tempClaudeRoot: string;
|
let tempClaudeRoot: string;
|
||||||
|
|
@ -124,17 +144,18 @@ liveDescribe('Member work sync Claude Stop hook live e2e', () => {
|
||||||
}
|
}
|
||||||
});
|
});
|
||||||
|
|
||||||
it(
|
async function runClaudeStopHookLiveScenario(
|
||||||
'launches a real Claude teammate, accepts its work-sync report, and ingests its Stop hook event',
|
scenario: ClaudeStopHookLiveScenario
|
||||||
async () => {
|
): Promise<void> {
|
||||||
const orchestratorCli = process.env.CLAUDE_AGENT_TEAMS_ORCHESTRATOR_CLI_PATH?.trim();
|
const orchestratorCli = process.env.CLAUDE_AGENT_TEAMS_ORCHESTRATOR_CLI_PATH?.trim();
|
||||||
expect(orchestratorCli).toBeTruthy();
|
expect(orchestratorCli).toBeTruthy();
|
||||||
await assertExecutable(orchestratorCli!);
|
await assertExecutable(orchestratorCli!);
|
||||||
|
|
||||||
const model = process.env.MEMBER_WORK_SYNC_CLAUDE_MODEL?.trim() || DEFAULT_MODEL;
|
const model = process.env.MEMBER_WORK_SYNC_CLAUDE_MODEL?.trim() || DEFAULT_MODEL;
|
||||||
const marker = `member-work-sync-claude-stop-live-${Date.now()}`;
|
const startedAt = Date.now();
|
||||||
|
const marker = `member-work-sync-claude-stop-live-${scenario.markerSuffix}-${startedAt}`;
|
||||||
const memberName = 'alice';
|
const memberName = 'alice';
|
||||||
teamName = `member-work-sync-claude-stop-${Date.now()}`;
|
teamName = `member-work-sync-claude-stop-${scenario.markerSuffix}-${startedAt}`;
|
||||||
const projectPath = path.join(tempDir, 'project');
|
const projectPath = path.join(tempDir, 'project');
|
||||||
await fs.mkdir(projectPath, { recursive: true });
|
await fs.mkdir(projectPath, { recursive: true });
|
||||||
await fs.writeFile(
|
await fs.writeFile(
|
||||||
|
|
@ -161,9 +182,9 @@ liveDescribe('Member work sync Claude Stop hook live e2e', () => {
|
||||||
queueQuietWindowMs: 500,
|
queueQuietWindowMs: 500,
|
||||||
// Native Claude teammates are registered by the real lead process, but in this
|
// Native Claude teammates are registered by the real lead process, but in this
|
||||||
// headless harness their bootstrap turn can finish before there is a durable
|
// headless harness their bootstrap turn can finish before there is a durable
|
||||||
// member process to prompt. The live assertion below still uses a real Claude
|
// member process to prompt. These live assertions still use a real Claude
|
||||||
// process, real MCP calls, and a real Stop hook payload; this seam keeps the
|
// process, real MCP calls, and real Stop hook payloads; this seam keeps the
|
||||||
// test focused on hook ingestion instead of tmux liveness.
|
// tests focused on hook ingestion instead of tmux liveness.
|
||||||
runtimeTurnSettledTargetResolver: {
|
runtimeTurnSettledTargetResolver: {
|
||||||
resolve: async (event) => {
|
resolve: async (event) => {
|
||||||
if (event.provider !== 'claude') {
|
if (event.provider !== 'claude') {
|
||||||
|
|
@ -181,9 +202,7 @@ liveDescribe('Member work sync Claude Stop hook live e2e', () => {
|
||||||
},
|
},
|
||||||
},
|
},
|
||||||
});
|
});
|
||||||
activeService.setTeamChangeEmitter((event: TeamChangeEvent) =>
|
activeService.setTeamChangeEmitter((event: TeamChangeEvent) => feature!.noteTeamChange(event));
|
||||||
feature!.noteTeamChange(event)
|
|
||||||
);
|
|
||||||
activeService.setRuntimeTurnSettledHookSettingsProvider((input) =>
|
activeService.setRuntimeTurnSettledHookSettingsProvider((input) =>
|
||||||
feature!.buildRuntimeTurnSettledHookSettings(input)
|
feature!.buildRuntimeTurnSettledHookSettings(input)
|
||||||
);
|
);
|
||||||
|
|
@ -242,33 +261,19 @@ liveDescribe('Member work sync Claude Stop hook live e2e', () => {
|
||||||
)
|
)
|
||||||
).resolves.toMatchObject({ mode: expect.any(Number) });
|
).resolves.toMatchObject({ mode: expect.any(Number) });
|
||||||
|
|
||||||
|
const taskPromptContext: ClaudeStopHookLiveScenarioContext = {
|
||||||
|
marker,
|
||||||
|
memberName,
|
||||||
|
teamName,
|
||||||
|
controlUrl: controlServer.baseUrl,
|
||||||
|
};
|
||||||
const task = await teamDataService.createTask(teamName, {
|
const task = await teamDataService.createTask(teamName, {
|
||||||
subject: `Member work sync Claude Stop hook live lease ${marker}`,
|
subject: `${scenario.subjectPrefix} ${marker}`,
|
||||||
owner: memberName,
|
owner: memberName,
|
||||||
startImmediately: true,
|
startImmediately: true,
|
||||||
prompt: [
|
prompt: scenario.buildTaskPromptLines(taskPromptContext).join('\n'),
|
||||||
`This is a live member-work-sync validation task. Marker: ${marker}.`,
|
|
||||||
'Do not edit files and do not complete this task.',
|
|
||||||
'Call task_start for this task.',
|
|
||||||
`Add one task comment containing exactly: ${marker}:still-working.`,
|
|
||||||
`Then call member_work_sync_status with teamName "${teamName}", memberName "${memberName}", and controlUrl "${controlServer.baseUrl}".`,
|
|
||||||
`Then call member_work_sync_report with teamName "${teamName}", memberName "${memberName}", controlUrl "${controlServer.baseUrl}", state "still_working", the exact agendaFingerprint and reportToken returned by member_work_sync_status, and the current task id if available.`,
|
|
||||||
'After that stop. Do not send a user-visible message.',
|
|
||||||
].join('\n'),
|
|
||||||
});
|
});
|
||||||
feature.noteTeamChange({ type: 'task', teamName, taskId: task.id });
|
feature.noteTeamChange({ type: 'task', teamName, taskId: task.id });
|
||||||
await activeService.sendMessageToTeam(
|
|
||||||
teamName,
|
|
||||||
[
|
|
||||||
`Live member-work-sync validation instruction. Marker: ${marker}.`,
|
|
||||||
`Use the board MCP tools as member "${memberName}" for this validation.`,
|
|
||||||
`Call task_get for taskId "${task.id}", then task_start.`,
|
|
||||||
`Add one task comment containing exactly: ${marker}:still-working.`,
|
|
||||||
`Then call member_work_sync_status with teamName "${teamName}", memberName "${memberName}", and controlUrl "${controlServer.baseUrl}".`,
|
|
||||||
`Then call member_work_sync_report with teamName "${teamName}", memberName "${memberName}", controlUrl "${controlServer.baseUrl}", state "still_working", the exact agendaFingerprint and reportToken returned by member_work_sync_status, and taskIds ["${task.id}"].`,
|
|
||||||
'After that stop. Do not complete the task. Do not send a user-visible message.',
|
|
||||||
].join('\n')
|
|
||||||
);
|
|
||||||
|
|
||||||
await waitUntil(async () => {
|
await waitUntil(async () => {
|
||||||
const status = await feature!.getStatus({ teamName: teamName!, memberName });
|
const status = await feature!.getStatus({ teamName: teamName!, memberName });
|
||||||
|
|
@ -280,18 +285,33 @@ liveDescribe('Member work sync Claude Stop hook live e2e', () => {
|
||||||
);
|
);
|
||||||
}, 30_000);
|
}, 30_000);
|
||||||
|
|
||||||
|
await activeService.sendMessageToTeam(
|
||||||
|
teamName,
|
||||||
|
scenario
|
||||||
|
.buildInstructionLines({
|
||||||
|
...taskPromptContext,
|
||||||
|
taskId: task.id,
|
||||||
|
})
|
||||||
|
.join('\n')
|
||||||
|
);
|
||||||
|
|
||||||
await waitUntil(async () => {
|
await waitUntil(async () => {
|
||||||
await feature!.replayPendingReports([teamName!]);
|
await feature!.replayPendingReports([teamName!]);
|
||||||
const status = await feature!.getStatus({ teamName: teamName!, memberName });
|
const [status, tasks] = await Promise.all([
|
||||||
if (status.report?.accepted && status.report.state === 'still_working') {
|
feature!.getStatus({ teamName: teamName!, memberName }),
|
||||||
return true;
|
new TeamTaskReader().getTasks(teamName!),
|
||||||
}
|
]);
|
||||||
const tasks = await new TeamTaskReader().getTasks(teamName!);
|
|
||||||
const currentTask = tasks.find((candidate) => candidate.id === task.id);
|
const currentTask = tasks.find((candidate) => candidate.id === task.id);
|
||||||
const hasMarkerComment = currentTask?.comments?.some((comment) =>
|
const expectedMarker = scenario.expectedMarkerText(marker);
|
||||||
comment.text.includes(`${marker}:still-working`)
|
const hasMarkerComment = currentTask?.comments?.some(
|
||||||
|
(comment) => comment.author === memberName && comment.text.includes(expectedMarker)
|
||||||
|
);
|
||||||
|
return Boolean(
|
||||||
|
hasMarkerComment &&
|
||||||
|
currentTask?.status === scenario.expectedTaskStatus &&
|
||||||
|
status.report?.accepted &&
|
||||||
|
status.report.state === scenario.expectedState
|
||||||
);
|
);
|
||||||
return Boolean(hasMarkerComment && status.report?.accepted);
|
|
||||||
}, 300_000, 2_000, async () =>
|
}, 300_000, 2_000, async () =>
|
||||||
formatMemberWorkSyncDiagnostics({
|
formatMemberWorkSyncDiagnostics({
|
||||||
feature: feature!,
|
feature: feature!,
|
||||||
|
|
@ -333,17 +353,84 @@ liveDescribe('Member work sync Claude Stop hook live e2e', () => {
|
||||||
feature.getStatus({ teamName, memberName }),
|
feature.getStatus({ teamName, memberName }),
|
||||||
feature.getMetrics({ teamName }),
|
feature.getMetrics({ teamName }),
|
||||||
]);
|
]);
|
||||||
expect(finalStatus.state).toBe('still_working');
|
expect(finalStatus.state).toBe(scenario.expectedState);
|
||||||
expect(finalStatus.report).toMatchObject({
|
expect(finalStatus.report).toMatchObject({
|
||||||
accepted: true,
|
accepted: true,
|
||||||
state: 'still_working',
|
state: scenario.expectedState,
|
||||||
});
|
});
|
||||||
|
if (scenario.expectedState === 'caught_up') {
|
||||||
|
expect(finalStatus.agenda.items).toHaveLength(0);
|
||||||
|
} else {
|
||||||
|
expect(finalStatus.agenda.items.some((item) => item.taskId === task.id)).toBe(true);
|
||||||
|
}
|
||||||
expect(metrics.recentEvents.some((event) => event.kind === 'report_accepted')).toBe(true);
|
expect(metrics.recentEvents.some((event) => event.kind === 'report_accepted')).toBe(true);
|
||||||
await expect(feature.dispatchDueNudges([teamName])).resolves.toMatchObject({
|
await expect(feature.dispatchDueNudges([teamName])).resolves.toMatchObject({
|
||||||
claimed: 0,
|
claimed: 0,
|
||||||
delivered: 0,
|
delivered: 0,
|
||||||
});
|
});
|
||||||
},
|
}
|
||||||
|
|
||||||
|
it(
|
||||||
|
'launches a real Claude teammate, accepts a still-working report, and ingests its Stop hook event',
|
||||||
|
async () =>
|
||||||
|
runClaudeStopHookLiveScenario({
|
||||||
|
markerSuffix: 'lease',
|
||||||
|
subjectPrefix: 'Member work sync Claude Stop hook live lease',
|
||||||
|
expectedState: 'still_working',
|
||||||
|
expectedTaskStatus: 'in_progress',
|
||||||
|
expectedMarkerText: (marker) => `${marker}:still-working`,
|
||||||
|
buildTaskPromptLines: ({ marker, memberName, teamName, controlUrl }) => [
|
||||||
|
`This is a live member-work-sync validation task. Marker: ${marker}.`,
|
||||||
|
'Do not edit files and do not complete this task.',
|
||||||
|
'Call task_start for this task.',
|
||||||
|
`Add one task comment containing exactly: ${marker}:still-working.`,
|
||||||
|
`Then call member_work_sync_status with teamName "${teamName}", memberName "${memberName}", and controlUrl "${controlUrl}".`,
|
||||||
|
`Then call member_work_sync_report with teamName "${teamName}", memberName "${memberName}", controlUrl "${controlUrl}", state "still_working", the exact agendaFingerprint and reportToken returned by member_work_sync_status, and the current task id if available.`,
|
||||||
|
'After that stop. Do not send a user-visible message.',
|
||||||
|
],
|
||||||
|
buildInstructionLines: ({ marker, memberName, teamName, controlUrl, taskId }) => [
|
||||||
|
`Live member-work-sync validation instruction. Marker: ${marker}.`,
|
||||||
|
`Use the board MCP tools as member "${memberName}" for this validation.`,
|
||||||
|
`Call task_get for taskId "${taskId}", then task_start.`,
|
||||||
|
`Add one task comment containing exactly: ${marker}:still-working.`,
|
||||||
|
`Then call member_work_sync_status with teamName "${teamName}", memberName "${memberName}", and controlUrl "${controlUrl}".`,
|
||||||
|
`Then call member_work_sync_report with teamName "${teamName}", memberName "${memberName}", controlUrl "${controlUrl}", state "still_working", the exact agendaFingerprint and reportToken returned by member_work_sync_status, and taskIds ["${taskId}"].`,
|
||||||
|
'After that stop. Do not complete the task. Do not send a user-visible message.',
|
||||||
|
],
|
||||||
|
}),
|
||||||
|
420_000
|
||||||
|
);
|
||||||
|
|
||||||
|
it(
|
||||||
|
'launches a real Claude teammate, completes work, reports caught-up, and ingests its Stop hook event',
|
||||||
|
async () =>
|
||||||
|
runClaudeStopHookLiveScenario({
|
||||||
|
markerSuffix: 'caught-up',
|
||||||
|
subjectPrefix: 'Member work sync Claude Stop hook live caught-up',
|
||||||
|
expectedState: 'caught_up',
|
||||||
|
expectedTaskStatus: 'completed',
|
||||||
|
expectedMarkerText: (marker) => `${marker}:completed`,
|
||||||
|
buildTaskPromptLines: ({ marker, memberName, teamName, controlUrl }) => [
|
||||||
|
`This is a live member-work-sync caught-up validation task. Marker: ${marker}.`,
|
||||||
|
'Do not edit files.',
|
||||||
|
'Call task_start for this task.',
|
||||||
|
`Add one task comment containing exactly: ${marker}:completed.`,
|
||||||
|
'Then call task_complete for this task.',
|
||||||
|
`Then call member_work_sync_status with teamName "${teamName}", memberName "${memberName}", and controlUrl "${controlUrl}".`,
|
||||||
|
`Then call member_work_sync_report with teamName "${teamName}", memberName "${memberName}", controlUrl "${controlUrl}", state "caught_up", the exact agendaFingerprint and reportToken returned by member_work_sync_status, and no taskIds.`,
|
||||||
|
'After that stop. Do not send a user-visible message.',
|
||||||
|
],
|
||||||
|
buildInstructionLines: ({ marker, memberName, teamName, controlUrl, taskId }) => [
|
||||||
|
`Live member-work-sync caught-up validation instruction. Marker: ${marker}.`,
|
||||||
|
`Use the board MCP tools as member "${memberName}" for this validation.`,
|
||||||
|
`Call task_get for taskId "${taskId}", then task_start.`,
|
||||||
|
`Add one task comment containing exactly: ${marker}:completed.`,
|
||||||
|
`Call task_complete for taskId "${taskId}".`,
|
||||||
|
`Then call member_work_sync_status with teamName "${teamName}", memberName "${memberName}", and controlUrl "${controlUrl}".`,
|
||||||
|
`Then call member_work_sync_report with teamName "${teamName}", memberName "${memberName}", controlUrl "${controlUrl}", state "caught_up", the exact agendaFingerprint and reportToken returned by member_work_sync_status, and no taskIds.`,
|
||||||
|
'After that stop. Do not send a user-visible message.',
|
||||||
|
],
|
||||||
|
}),
|
||||||
420_000
|
420_000
|
||||||
);
|
);
|
||||||
});
|
});
|
||||||
|
|
|
||||||
Loading…
Reference in a new issue