@warren-dean/code-kit 0.4.1 → 0.6.0

package/hooks/mod/register.test.ts

@@ -6,6 +6,7 @@ import { expect, mock, test } from 'claude-code/testing';
 const PANE = 'code-kit-lanes';
 const APPROVE = 'code-kit-approve';
 const RESULT = 'code-kit-result';
+const SETTINGS = 'code-kit-settings';
 
 /** A project as the stubs report it; tests change it to move the world on. */
 function project() {
@@ -68,7 +69,8 @@ function project() {
     taken: [] as string[], // command names another command already holds
     stops: [] as any[],
     refs: ' aaa refs/heads/main\n bbb refs/heads/web/st-4\n',
-    agents: [] as { type: string; status: string }[],
+    agents: [] as { type: string; status: string; id?: string; description?: string }[],
+    limits: [] as { kind: string; percentUsed: number }[],
     approveExit: 0,
     mergeExit: 0,
     answer: 'Merge',
@@ -78,6 +80,18 @@ function project() {
     prompts: [] as string[],
     acts: [] as string[][],
     runs: 0,
+    settings: [
+      {
+        key: 'autonomy',
+        value: 'autonomous',
+        default: 'autonomous',
+        about: 'How the lead loop acts',
+      },
+      { key: 'hold.minutes', value: 2, default: 2, about: 'How long a call is held' },
+      { key: 'agents.reviewer.on', value: false, default: false, about: 'A background reviewer' },
+      { key: 'agents.reviewer.model', value: null, default: null, about: "The reviewer's model" },
+    ] as any[],
+    setExit: 0,
   };
 }
 type World = ReturnType<typeof project>;
@@ -119,6 +133,29 @@ function stub(on: any, w: World) {
         ? ran(1, '', 'Nothing was merged. code-kit verify found 1 problem(s) on web/st-4')
         : ran(0, 'Merged web/st-4 into main.');
     }
+    if (sub === 'settings' && argv[3] === 'set') {
+      w.acts.push([...argv.slice(2)]);
+      if (w.setExit)
+        return ran(
+          1,
+          '',
+          '"harness.hold.minutes" must be a number of minutes from 0 (off) to 30\nNothing was changed.',
+        );
+      const row = w.settings.find((s) => s.key === argv[4]);
+      const raw = argv[5];
+      row.value =
+        raw === 'null'
+          ? null
+          : raw === 'true'
+            ? true
+            : raw === 'false'
+              ? false
+              : /^\d+$/.test(raw)
+                ? Number(raw)
+                : raw;
+      return ran(0, `Set harness.${argv[4]} to ${raw}.`);
+    }
+    if (sub === 'settings') return ran(0, JSON.stringify(w.settings));
     return ran(1, '', `unexpected ${argv.join(' ')}`);
   });
   on('session.start', () => ({ cwd: '/work' }));
@@ -135,6 +172,19 @@ function stub(on: any, w: World) {
     w.refusal ? { deny: w.refusal } : { result: { stdout: '', stderr: '', interrupted: false } },
   );
   on('agent.list', () => ({ value: w.agents }));
+  on('session.usage', () => ({ value: { startedAt: 0, context: {}, rateLimits: w.limits } }));
+  on('session.measure', ($: any, e: any) => ({ changed: e.changed }));
+  // A model request, answered with the usage the test gave it.
+  on('turn.step', async function* ($: any, e: any) {
+    return {
+      turnId: e.turnId,
+      index: e.index,
+      answer: '',
+      toolUses: [],
+      stopReason: 'end_turn',
+      usage: e.messageCount === 0 ? null : { model: e.model, ...stepUsage(e.messageCount) },
+    };
+  } as any);
   on('fs.stat', () => ({ value: { kind: 'file', size: 1, mtimeMs: 1, isLink: false } }));
   on('fs.list', () => ({ value: [] }));
   on('session.cwd', () => ({ value: '/work' }));
@@ -163,6 +213,30 @@ function stub(on: any, w: World) {
   });
 }
 
+/** A request of n thousand tokens, split as a model request's usage is. */
+function stepUsage(thousands: number) {
+  const n = thousands * 1000;
+  return {
+    input_tokens: n * 0.3,
+    cache_creation_input_tokens: n * 0.1,
+    output_tokens: n * 0.1,
+    cache_read_input_tokens: n * 0.5,
+  };
+}
+
+/** One model request by an agent (or the lead) of so many thousand tokens, run to its end. */
+async function request($: any, agentId: string | undefined, thousands: number) {
+  const s = $.turn.step({
+    turnId: `t-${agentId ?? 'lead'}`,
+    index: 0,
+    model: 'claude-sonnet-5-5',
+    messageCount: thousands, // the stub beneath answers with this many thousand tokens
+    ...(agentId ? { agentId } : {}),
+  });
+  for await (const _ of s);
+  return s.result;
+}
+
 /** A session started on the project, with the mock clock. */
 async function start($: any, on: any, w: World) {
   stub(on, w);
@@ -336,7 +410,9 @@ test('ACT-1 and BAND-3 approving with the prefilled reason grants it for the lan
   const confirm = await pane($, APPROVE);
   const reason = 'Approve dayjs for the web lane: npm install dayjs';
   expect((await confirm.find({ key: 'approve-reason' }))?.props.value).toBe(reason);
-  expect(await confirm.find({ type: 'Text', text: /for 60 minutes/ })).toBeDefined();
+  expect(
+    await confirm.find({ type: 'Text', text: /until the install is committed \(at most 7 days\)/ }),
+  ).toBeDefined();
   await press($, 'approve-confirm', APPROVE);
   expect(w.acts).toEqual([
     ['approve', 'dep-dayjs', '--lane', 'web', '--reason', reason, '--via', 'pane'],
@@ -558,7 +634,7 @@ test('CARD-1 a refused write is drawn as a card, its text a press away', async (
 test('CARD-1 a refusal an approval would allow has Approve…, which opens the confirmation', async ($, on) => {
   const w = project();
   w.refusal =
-    'Blocked: a new dependency (dayjs) needs a person\'s approval. Write a short change request for the lead.\n  ! echo "<what you are approving>" > .claude/approvals/web/dep-dayjs\n(a person runs it; it allows installing it for the web lane, for 60 minutes)\nCommand: npm install dayjs';
+    'Blocked: a new dependency (dayjs) needs a person\'s approval. Write a short change request for the lead.\n  ! echo "<what you are approving>" > .claude/approvals/web/dep-dayjs\n(a person runs it; it allows installing it for the web lane, until the install is committed, at most 7 days)\nCommand: npm install dayjs';
   await start($, on, w);
   await $.tool.call({ tool: 'Bash', tool_use_id: 'tu-2', command: 'npm install dayjs' } as any);
   const ui = await toolResult($, 'tu-2', 'Bash');
@@ -622,7 +698,7 @@ test('CARD-4 outside a code-kit project, /approvals and /verify-branch say so',
 });
 
 const install =
-  'Blocked: a new dependency (dayjs) needs a person\'s approval. Write a short change request for the lead.\n  ! echo "<what you are approving>" > .claude/approvals/dep-dayjs\n(a person runs it; it allows installing it for any agent, for 60 minutes)\nCommand: npm install dayjs';
+  'Blocked: a new dependency (dayjs) needs a person\'s approval. Write a short change request for the lead.\n  ! echo "<what you are approving>" > .claude/approvals/dep-dayjs\n(a person runs it; it allows installing it for any agent, until the install is committed, at most 7 days)\nCommand: npm install dayjs';
 
 test('CARD-1 a group holding a refusal unfolds, so its row can be the card', async ($, on) => {
   const w = project();
@@ -712,3 +788,228 @@ test(
     await ui.unmount();
   },
 );
+
+// --- the harness settings -------------------------------------------------------------------------
+
+const harness = ($: any) => $.command.run({ command: 'harness', args: '' });
+
+test('SET-2 /harness opens the settings, each with its value, and /harness again closes it', async ($, on) => {
+  const w = project();
+  await start($, on, w);
+  await harness($);
+  expect(w.opened).toContain(SETTINGS);
+  const ui = await pane($, SETTINGS);
+  expect(await ui.find({ type: 'Text', text: 'hold.minutes' })).toBeDefined();
+  expect((await ui.find({ key: 'set-hold.minutes' }))?.props.value).toBe('2');
+  expect((await ui.find({ key: 'set-autonomy' }))?.props.value).toBe('autonomous');
+  await ui.unmount();
+  await harness($);
+  expect(w.closed).toContain(SETTINGS);
+});
+
+test('SET-2 a change asks for a reason, then is written by the CLI as the pane', async ($, on) => {
+  const w = project();
+  await start($, on, w);
+  await harness($);
+  const ui = await pane($, SETTINGS);
+  await $.ui.select({
+    plugin: 'code-kit',
+    key: 'set-autonomy',
+    value: 'propose',
+    requestId: SETTINGS,
+  });
+  expect(w.acts).toEqual([]);
+  expect(await ui.find({ type: 'Text', text: /Set autonomy to propose\?/ })).toBeDefined();
+  await $.ui.input({ plugin: 'code-kit', key: 'settings-reason', text: '   ' });
+  expect(w.acts).toEqual([]);
+  expect(await ui.find({ type: 'Text', text: /Give a reason/ })).toBeDefined();
+  await $.ui.input({ plugin: 'code-kit', key: 'settings-reason', text: 'watch it for a week' });
+  expect(w.acts).toEqual([
+    ['settings', 'set', 'autonomy', 'propose', '--reason', 'watch it for a week', '--via', 'pane'],
+  ]);
+  expect(await ui.find({ key: 'settings-confirm' })).toBeUndefined();
+  expect((await ui.find({ key: 'set-autonomy' }))?.props.value).toBe('propose');
+  await ui.unmount();
+});
+
+test('SET-2 a number is typed; one the CLI refuses shows why and changes nothing', async ($, on) => {
+  const w = project();
+  w.setExit = 1;
+  await start($, on, w);
+  await harness($);
+  const ui = await pane($, SETTINGS);
+  await $.ui.input({ plugin: 'code-kit', key: 'set-hold.minutes', text: '45' });
+  await $.ui.input({ plugin: 'code-kit', key: 'settings-reason', text: 'longer' });
+  expect(await ui.find({ type: 'Text', text: /from 0 \(off\) to 30/ })).toBeDefined();
+  expect(w.settings[1].value).toBe(2);
+  await press($, 'settings-cancel', SETTINGS);
+  expect(await ui.find({ key: 'settings-confirm' })).toBeUndefined();
+  await ui.unmount();
+});
+
+test("SET-2 the reviewer's model can go back to the session's", async ($, on) => {
+  const w = project();
+  w.settings[3].value = 'haiku';
+  await start($, on, w);
+  await harness($);
+  const ui = await pane($, SETTINGS);
+  expect(await ui.find({ type: 'Text', text: /default the session's/ })).toBeDefined();
+  await $.ui.select({
+    plugin: 'code-kit',
+    key: 'set-agents.reviewer.model',
+    value: '',
+    requestId: SETTINGS,
+  });
+  await press($, 'settings-ok', SETTINGS);
+  expect(w.acts).toEqual([]);
+  await $.ui.input({ plugin: 'code-kit', key: 'settings-reason', text: 'same as the lead' });
+  expect(w.acts[0].slice(0, 4)).toEqual(['settings', 'set', 'agents.reviewer.model', 'null']);
+  await ui.unmount();
+});
+
+test('CARD-4 outside a code-kit project, /harness says so and opens nothing', async ($, on) => {
+  const w = project();
+  w.check.exists = false;
+  w.check.valid = false;
+  await start($, on, w);
+  const said = await harness($);
+  expect(String(said?.text)).toMatch(/code-kit/);
+  expect(w.opened).not.toContain(SETTINGS);
+});
+
+// --- usage per lane ---------------------------------------------------------------------------------
+
+const working = (id: string, type: string, description: string) => ({
+  id,
+  type,
+  description,
+  status: 'running',
+});
+
+test("USE-1 a lane agent's requests show on its lane's row and its story", async ($, on) => {
+  const w = project();
+  w.agents = [working('a1', 'web-engineer', 'Build ST-4: the booking form')];
+  const clock = await start($, on, w);
+  await lanes($);
+  await request($, 'a1', 120);
+  await request($, 'a1', 80);
+  await clock.advance(2000);
+  const ui = await pane($, PANE);
+  expect((await ui.find({ key: 'lane-web' }))?.text).toContain('200k');
+  expect((await ui.find({ key: 'usage-ST-4' }))?.text).toContain('200k');
+  expect((await ui.find({ key: 'lane-api' }))?.text).not.toMatch(/\d+k/);
+  await ui.unmount();
+});
+
+test("USE-1 the lead's requests are the lead's; an agent's brief names its story", async ($, on) => {
+  const w = project();
+  w.agents = [working('a2', 'core-engineer', 'ST-9 pricing rules')];
+  await start($, on, w);
+  await lanes($);
+  await request($, undefined, 30);
+  await request($, 'a2', 50);
+  const ui = await pane($, PANE);
+  expect((await ui.find({ key: 'lane-lead' }))?.text).toContain('30k');
+  expect((await ui.find({ key: 'lane-core' }))?.text).toContain('50k');
+  expect((await ui.find({ key: 'usage-ST-9' }))?.text).toContain('50k');
+  await ui.unmount();
+});
+
+test("USE-2 the background agents' share and the plan's 5-hour use", async ($, on) => {
+  const w = project();
+  w.limits = [{ kind: 'five_hour', percentUsed: 62 }];
+  w.agents = [
+    working('a1', 'web-engineer', 'ST-4 booking form'),
+    working('r1', 'code-reviewer', 'Review web/st-4'),
+  ];
+  await start($, on, w);
+  await lanes($);
+  await request($, 'a1', 150);
+  await request($, 'r1', 50);
+  const ui = await pane($, PANE);
+  expect(await ui.find({ type: 'Text', text: 'Background agents  50k (25%)' })).toBeDefined();
+  expect(await ui.find({ type: 'Text', text: '62% of the 5-hour window' })).toBeDefined();
+  expect(await ui.find({ type: 'Text', text: 'This session  200k tokens' })).toBeDefined();
+  await ui.unmount();
+});
+
+test('USE-3 a story past three times the median of three finished ones is flagged in the band', async ($, on) => {
+  const w = project();
+  w.stories = [
+    ...['ST-1', 'ST-2', 'ST-3'].map((id) => ({
+      id,
+      title: id,
+      lane: 'api',
+      state: 'done',
+      branch: `api/${id.toLowerCase()}`,
+      waitingOn: [],
+    })),
+    ...w.stories,
+  ];
+  w.agents = ['ST-1', 'ST-2', 'ST-3', 'ST-4'].map((id, i) =>
+    working(`a${i}`, i < 3 ? 'api-engineer' : 'web-engineer', `${id} work`),
+  );
+  await start($, on, w);
+  await request($, 'a0', 190);
+  await request($, 'a1', 200);
+  await request($, 'a2', 210);
+  await request($, 'a3', 500);
+  let ui = await bandUi($);
+  expect(await ui.find({ type: 'Text', text: /has used/ })).toBeUndefined();
+  await ui.unmount();
+  await request($, 'a3', 200);
+  ui = await bandUi($);
+  expect(await ui.find({ type: 'Text', text: 'ST-4 has used 3.5× the usual' })).toBeDefined();
+  await ui.unmount();
+});
+
+test('USE-3 with fewer than three finished, nothing is flagged', async ($, on) => {
+  const w = project();
+  w.stories = [
+    { id: 'ST-1', title: 'a', lane: 'api', state: 'done', branch: 'api/st-1', waitingOn: [] },
+    ...w.stories,
+  ];
+  w.agents = [working('a0', 'api-engineer', 'ST-1'), working('a1', 'web-engineer', 'ST-4')];
+  await start($, on, w);
+  await request($, 'a0', 10);
+  await request($, 'a1', 900);
+  const ui = await bandUi($);
+  expect(await ui.find({ type: 'Text', text: /has used/ })).toBeUndefined();
+  await ui.unmount();
+});
+
+test('USE-4 past the pause point the band says background agents are paused, and below it not', async ($, on) => {
+  const w = project();
+  await start($, on, w);
+  const measure = (percentUsed: number) =>
+    $.session.measure({
+      context: {} as any,
+      rateLimits: [{ kind: 'five_hour', percentUsed }],
+      changed: ['rateLimits'],
+    });
+  await measure(82);
+  let ui = await bandUi($);
+  expect(
+    await ui.find({ type: 'Text', text: 'background agents paused: plan at 82%' }),
+  ).toBeDefined();
+  await ui.unmount();
+  await measure(79);
+  ui = await bandUi($);
+  expect(await ui.find({ type: 'Text', text: /paused/ })).toBeUndefined();
+  await ui.unmount();
+});
+
+test('USE-4 the pause point is the harness setting', async ($, on) => {
+  const w = project();
+  (w.check as any).harness = { background: { pauseAtPercent: 90 } };
+  w.limits = [{ kind: 'five_hour', percentUsed: 85 }];
+  await start($, on, w);
+  let ui = await bandUi($);
+  expect(await ui.find({ type: 'Text', text: /paused/ })).toBeUndefined();
+  await ui.unmount();
+  w.limits = [{ kind: 'five_hour', percentUsed: 91 }];
+  await $.session.measure({ context: {} as any, rateLimits: w.limits, changed: ['rateLimits'] });
+  ui = await bandUi($);
+  expect(await ui.find({ type: 'Text', text: /plan at 91%/ })).toBeDefined();
+  await ui.unmount();
+});