Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
Python agentaichatgptdeepseekfine-tuninggemmaimage-generationllamallmllmsopenaipythonqwenreinforcement-learningself-hostedstable-diffusiontext-to-speechttsuiunsloth
Studio Playwright chat_ui: condition waits, per-step budgets, 25 s less in the theme step (#11983)
Daniel Han6c1687db4e43
diff --git a/tests/studio/playwright_chat_ui.py b/tests/studio/playwright_chat_ui.pyindex 03132d803ff..c4c3580c537 100644--- a/tests/studio/playwright_chat_ui.py+++ b/tests/studio/playwright_chat_ui.py@@ -27,8 +27,13 @@ from _playwright_robust import ( # noqa: E402is_benign_console_error,is_benign_page_error,recover_or_replace_page as _robust_recover_or_replace_page,+ report_failing_step,robust_evaluate,+ step_budget_s,+ wait_for_first,wait_for_health,+ wait_for_settled,+ wait_until,click_forced,)@@ -93,6 +98,20 @@ WALL_TIMEOUT_S = float(# .github/scripts/run-studio-permission-browser.sh.TOTAL_TIMEOUT_S = float(os.environ.get("STUDIO_UI_TOTAL_TIMEOUT_S", "0")) or None+# Per-step ceilings. A step that overruns its own stops the run there, named, instead of+# every later step waiting out its own timeouts first. A step that waits on the model gets+# WALL_TIMEOUT_S, the budget each of its waits already had; one that only drives the UI+# gets UI_STEP_BUDGET_S, generous next to the few seconds it takes on a hosted runner.+# Both stretch with STUDIO_UI_TURN_TIMEOUT_MS, the knob the slow lanes already raise+# (macOS triples it), and with STUDIO_PW_STEP_BUDGET_SCALE.+_SLOW_LANE = max(1.0, TURN_TIMEOUT_MS / 180_000)+UI_STEP_BUDGET_S = step_budget_s(180 * _SLOW_LANE)+MODEL_STEP_BUDGET_S = step_budget_s(WALL_TIMEOUT_S)+# For the steps that retry themselves (change-password, composer mount, re-login): each+# attempt restarts the inactivity budget, as it always has, and the step has no ceiling of+# its own, so no attempt the loop allows is cut short. Failing still names the step.+NO_STEP_CEILING = 0+_n = [0]_watchdog = None # armed below; everything above it runs before there is one@@ -104,9 +123,11 @@ def wall_kick():_watchdog.kick()-def step(s):+def step(s, budget_s = None):+ """Start step `s`; it may run `budget_s` (default UI_STEP_BUDGET_S) before the run stops."""print(f"[ui] STEP {s}", flush = True)- wall_kick()+ if _watchdog is not None:+ _watchdog.begin_step(s, UI_STEP_BUDGET_S if budget_s is None else budget_s)def info(s):@@ -411,7 +432,8 @@ def exercise_permission_mode_controls(page, shoot):expect(pill).to_have_attribute("data-variant", "danger")active_icon = pill.locator(".composer-pill-glyph > :first-child")pill.hover()- page.wait_for_timeout(200)+ # Read the opacity once the hover transition has finished, not at a fixed delay into it.+ wait_for_settled(active_icon)icon_opacity = float(active_icon.evaluate("el => getComputedStyle(el).opacity"))if icon_opacity < 0.5:fail(f"Full access icon disappeared on hover (opacity={icon_opacity})")@@ -623,7 +645,8 @@ def exercise_floating_monitor_geometry(page):page.mouse.down()page.mouse.move(end_x, end_y, steps = 10)page.mouse.up()- page.wait_for_timeout(100)+ # The callers measure the panel next: wait for it to stop moving, not for 100 ms.+ wait_for_settled(monitor)def drag_monitor_to(x, y):box = monitor_handle.bounding_box()@@ -851,6 +874,7 @@ with sync_playwright() as p:info = info,total_deadline_s = TOTAL_TIMEOUT_S,)+ report_failing_step(_watchdog, label = "ui")# Pre-flight: macos-14 can surface a 200 /api/health while the auth DB is still migrating;# this 30s probe catches that gap before we sink 60s into a change-password timeout.wait_for_health(BASE, timeout = 30.0, info = info)@@ -936,7 +960,7 @@ with sync_playwright() as p:# Bootstrap state pre-seeds the current password; we enter the# new password twice and submit -- the user's first-run experience.# ─────────────────────────────────────────────────────- step("change-password through UI (Setup your account)")+ step("change-password through UI (Setup your account)", NO_STEP_CEILING)# Settle the network before touching the form: a late bootstrap poll can rerender the page (dropping# #new-password) mid-test. The whole goto/wait/fill/submit sequence is wrapped in a 3-attempt retry with a fresh# page/reload between tries so a mid-try rerender doesn't poison the next.@@ -1022,7 +1046,7 @@ with sync_playwright() as p:# ─────────────────────────────────────────────────────# 2. Chat surface mounts, default model surface is visible.# ─────────────────────────────────────────────────────- step("wait for composer to mount")+ step("wait for composer to mount", NO_STEP_CEILING)# After change-password the router rebuilds login -> chat shell; on macos-14 racing straight into wait_for() either# burns the timeout or crashes the renderer mid-mount. Settle network first, then wait_for with one recovery cycle# on failure.@@ -1173,7 +1197,7 @@ with sync_playwright() as p:# ─────────────────────────────────────────────────────# 3. Trigger model load via the same endpoint the picker uses.# ─────────────────────────────────────────────────────- step("load GGUF via /api/inference/load (uses session cookie)")+ step("load GGUF via /api/inference/load (uses session cookie)", MODEL_STEP_BUDGET_S)# AbortSignal-bounded: macos-14 has been seen wedging on this fetch. The 3-min budget is generous for a# cold-cache load; a wedge fails cleanly instead of forcing a 30-min runner cancel.load_resp = evaluate_fetch(@@ -1223,12 +1247,13 @@ with sync_playwright() as p:soft_fail("model picker button not found")else:picker_btn.click()- page.wait_for_timeout(500)+ # The popover mounts its search input when it opens; wait for that, not 500 ms.+ search = wait_for_first(+ page.get_by_placeholder(re.compile(r"Search.*models?", re.I)),+ timeout_ms = 10_000,+ )shoot("03c-model-picker-open")- search = page.get_by_placeholder(- re.compile(r"Search.*models?", re.I),- ).first- if search.count() == 0:+ if search is None:soft_fail("model picker search input not found")else:# "qwen" then "llama" popover text must DIFFER, proving the typeahead actually filters (else an@@ -1244,15 +1269,37 @@ with sync_playwright() as p:}""",)- search.fill("qwen")- page.wait_for_timeout(800)- qwen_text = picker_visible_text()+ # The query is debounced, then filters locally and searches the Hub. Instead of+ # reading at a fixed 800 ms, wait for the popover text to move off what it showed+ # before this query and hold still for a few polls, then read it. A typeahead that+ # ignores its input never moves, the wait gives up, and the comparison below+ # reports it exactly as before.+ def picker_text_after(query, previous):+ search.fill(query)+ try:+ page.wait_for_function(+ """([previous]) => {+ const el = document.querySelector(+ '[role="dialog"], [role="listbox"], [role="menu"]'+ );+ const text = el ? (el.innerText || '').trim() : '';+ const s = (window.__pickerText ||= { text: null, same: 0 });+ s.same = text === s.text ? s.same + 1 : 0;+ s.text = text;+ return text !== previous && s.same >= 3;+ }""",+ arg = [previous],+ polling = 100,+ timeout = 10_000,+ )+ except Exception as exc:+ info(f"WARN picker text did not change for {query!r}: {type(exc).__name__}")+ return picker_visible_text()++ qwen_text = picker_text_after("qwen", picker_visible_text())shoot("03d-model-picker-search-qwen")search.fill("")- page.wait_for_timeout(300)- search.fill("llama")- page.wait_for_timeout(800)- llama_text = picker_visible_text()+ llama_text = picker_text_after("llama", qwen_text)shoot("03e-model-picker-search-llama")if qwen_text and llama_text and qwen_text == llama_text:soft_fail(@@ -1262,7 +1309,12 @@ with sync_playwright() as p:else:info("OK search bar filtered (qwen text != llama text)")page.keyboard.press("Escape")- page.wait_for_timeout(300)+ try:+ page.get_by_placeholder(re.compile(r"Search.*models?", re.I)).first.wait_for(+ state = "hidden", timeout = 10_000+ )+ except Exception:+ pass # best-effort, as the fixed pause was: the next step opens its own surface# ─────────────────────────────────────────────────────# 4. A follow-up submitted 100 ms after a normal send must queue behind it.@@ -1370,7 +1422,7 @@ with sync_playwright() as p:shoot(f"04-turn-{idx}-still-streaming")raise- step("rapid submit: 100 ms follow-up queues behind the first turn")+ step("rapid submit: 100 ms follow-up queues behind the first turn", MODEL_STEP_BUDGET_S)rapid_bubbles_before = _bubble_count()composer_form = page.locator('form:has(textarea[aria-label="Message input"])').first# How long a reply takes is not ours to decide: sampling settings, whatever GGUF_REPO points at and an early EOS all@@ -1565,7 +1617,7 @@ with sync_playwright() as p:info("OK 100 ms follow-up waited and both assistant turns completed")for i, p_ in enumerate(prompts, start = 1):- step(f"turn {i}: {p_!r}")+ step(f"turn {i}: {p_!r}", MODEL_STEP_BUDGET_S)send_and_wait(p_, i)shoot("04-after-five-turns")@@ -1590,10 +1642,12 @@ with sync_playwright() as p:# ─────────────────────────────────────────────────────# 5. Regenerate the last assistant turn.# ─────────────────────────────────────────────────────- step("regenerate last assistant turn")+ step("regenerate last assistant turn", MODEL_STEP_BUDGET_S)last_assistant = page.locator('[data-role="assistant"]').lastlast_assistant.hover()- page.wait_for_timeout(400)+ # The newest reply keeps its action bar mounted (autohide="not-last"), so wait for the bar+ # itself rather than a fixed 400 ms after the hover.+ wait_for_first(last_assistant.locator(".aui-assistant-action-bar-root"), timeout_ms = 10_000)# Exclude disabled controls: the picker's new disabled "Reload model" button also matches and sorts first, so# .first would target it.regen_btn = (@@ -1627,7 +1681,7 @@ with sync_playwright() as p:# ─────────────────────────────────────────────────────extra = ["Reply with: yes", "Reply with: no"]for j, p_ in enumerate(extra, start = 1):- step(f"extra turn {j}: {p_!r}")+ step(f"extra turn {j}: {p_!r}", MODEL_STEP_BUDGET_S)before_count = len(page.locator('[data-role="assistant"]').all())send_and_wait(p_, before_count + 1)shoot("06-after-extra-turns")@@ -1665,7 +1719,15 @@ with sync_playwright() as p:continuebefore = toggle.get_attribute("aria-label") or ""toggle.click()- page.wait_for_timeout(200)+ toggle_now = page.locator(+ f'button[aria-label="Disable {feature}"], button[aria-label="Enable {feature}"]'+ ).first+ # Wait for the label to flip rather than 200 ms; a toggle that never flips falls+ # through to the same WARN as before.+ try:+ expect(toggle_now).not_to_have_attribute("aria-label", before, timeout = 5_000)+ except AssertionError:+ passafter = (page.locator(f'button[aria-label="Disable {feature}"], button[aria-label="Enable {feature}"]'@@ -1680,9 +1742,9 @@ with sync_playwright() as p:page.locator(f'button[aria-label="Disable {feature}"], button[aria-label="Enable {feature}"]').first.click()+ expect(toggle_now).to_have_attribute("aria-label", before, timeout = 5_000)except Exception:pass- page.wait_for_timeout(200)shoot("07-toggles-cycled")# ─────────────────────────────────────────────────────@@ -1692,11 +1754,12 @@ with sync_playwright() as p:if cfg_open.count() > 0:step("Configuration sheet: drive Temperature + Top P + extras")cfg_open.click()- page.wait_for_timeout(500)+ # Count the sliders once the sheet has rendered them, not 500 ms after the click.+ sliders = page.locator('[role="slider"]')+ wait_for_first(sliders, timeout_ms = 10_000)shoot("08-config-open")# Walk every Radix slider (role="slider") by index, focus it, press Home (-> min) for deterministic state; a# locked slider surfaces an error here.- sliders = page.locator('[role="slider"]')n_sliders = sliders.count()info(f"configuration sheet exposes {n_sliders} slider(s)")for idx in range(n_sliders):@@ -1705,7 +1768,9 @@ with sync_playwright() as p:s.scroll_into_view_if_needed()s.focus()page.keyboard.press("Home") # -> min- page.wait_for_timeout(80)+ expect(s).to_have_attribute(+ "aria-valuenow", s.get_attribute("aria-valuemin") or "", timeout = 5_000+ )except Exception as exc:info(f" slider[{idx}] focus/Home failed: {exc!r}")shoot("09-config-all-min")@@ -1716,7 +1781,10 @@ with sync_playwright() as p:close_btn.click()else:page.keyboard.press("Escape")- page.wait_for_timeout(300)+ try:+ expect(sliders).to_have_count(0, timeout = 10_000)+ except AssertionError:+ pass # best-effort, as the fixed pause wasdef read_chat_typography():"""Read message typography after a user-driven theme transition."""@@ -1834,7 +1902,8 @@ with sync_playwright() as p:step("theme toggle x3 with computed-color assertion")observed = []typography_states = []- for cycle in range(3):++ def wait_menu_closed(timeout_ms):try:page.wait_for_function("""() => {@@ -1844,11 +1913,15 @@ with sync_playwright() as p:// close animation; treat that as already gone.return m.getAttribute('data-state') === 'closed';}""",- timeout = 7_000,+ timeout = timeout_ms,)except Exception:pass- page.wait_for_timeout(250)++ for cycle in range(3):+ # The previous cycle's menu is closed (or closing) when this returns, which is what+ # the extra 250 ms pause after it was standing in for.+ wait_menu_closed(7_000)# Retry once (after Escape to clear stray popups) if the first click is silently swallowed# mid-view-transition.opened = False@@ -1868,7 +1941,7 @@ with sync_playwright() as p:breakexcept Exception:page.keyboard.press("Escape")- page.wait_for_timeout(300)+ wait_menu_closed(5_000)if not opened:soft_fail(f"theme cycle {cycle + 1}: account menu didn't open")break@@ -1880,6 +1953,9 @@ with sync_playwright() as p:page.keyboard.press("Escape")soft_fail(f"theme cycle {cycle + 1}: theme menuitem missing")break+ was_dark = robust_evaluate(+ page, "() => document.documentElement.classList.contains('dark')"+ )# Click with fallbacks: a small CI viewport can push the item off-screen (force=True still needs it in# viewport). Fall back to scroll-into-view, then a synthetic evaluate() .click() that skips Playwright's# viewport check.@@ -1897,14 +1973,43 @@ with sync_playwright() as p:breakexcept Exception as exc:click_err = exc+ # Kept: a short pause between two click strategies, not a wait for any state.page.wait_for_timeout(200)if click_err is not None:page.keyboard.press("Escape")soft_fail(f"theme cycle {cycle + 1}: theme menuitem click failed ({click_err!r})")break# Settle. The ".dark" class on <html> is the ground truth (theme-store toggles only that); don't gate on- # ".light".- page.wait_for_timeout(700)+ # ".light". Instead of 700 ms: wait for it to flip, for every CSS transition the flip started to finish+ # (font-weight is animatable, and a read mid-transition reports the old theme's weight), and for the body+ # colour and message typography to hold for a few frames. A toggle that does not flip falls through to+ # the same polarity check as before.+ try:+ page.wait_for_function(+ """([wasDark]) => {+ const root = document.documentElement;+ const msg = document.querySelector(+ '.aui-assistant-message-root, .aui-user-message-root'+ );+ const ms = msg ? getComputedStyle(msg) : null;+ const sig = [+ getComputedStyle(document.body).backgroundColor,+ ms ? ms.fontWeight + '/' + ms.letterSpacing : '',+ ].join('|');+ const transitioning = document.getAnimations().some(+ (a) => a instanceof CSSTransition && a.playState === 'running'+ );+ const s = (window.__themeSettle ||= { sig: null, same: 0 });+ s.same = !transitioning && sig === s.sig ? s.same + 1 : 0;+ s.sig = sig;+ return root.classList.contains('dark') !== wasDark && s.same >= 3;+ }""",+ arg = [was_dark],+ polling = "raf",+ timeout = 10_000,+ )+ except Exception as exc:+ info(f"WARN theme cycle {cycle + 1}: .dark did not flip: {type(exc).__name__}")bg = robust_evaluate(page,"""() => {@@ -1923,6 +2028,12 @@ with sync_playwright() as p:typography_states.append(typography)shoot(f"10-theme-cycle-{cycle + 1}")info(f" cycle {cycle + 1}: dark={bg['isDark']} body bg={bg['bg']!r}")+ # The theme item keeps the menu open on purpose (preventDefault in app-sidebar.tsx). Close it here, as a+ # user would; left open, the next cycle waited up to 7 s for a close that never came, and its first click on+ # the account button only closed the menu, costing a failed open wait and a retry. Measured: about 10 s+ # per cycle before, 2 s after.+ page.keyboard.press("Escape")+ wait_menu_closed(5_000)# Across cycles we should see both a near-white (light) and a near-black (dark) body bg; one polarity means the# toggle stuck.rgbs = [parse_rgb(o["bg"]) for o in observed if parse_rgb(o["bg"])]@@ -1981,17 +2092,17 @@ with sync_playwright() as p:more_btn = page.get_by_role("button", name = re.compile(r"^\s*More\s*$", re.I)).firstif more_btn.count() > 0:more_btn.hover()- page.wait_for_timeout(500)- item = page.get_by_role("menuitem", name = re.compile(label, re.I)).first- if item.count() == 0:+ items = page.get_by_role("menuitem", name = re.compile(label, re.I))+ item = wait_for_first(items, timeout_ms = 5_000)+ if item is None:click_forced(more_btn)- page.wait_for_timeout(500)- item = page.get_by_role("menuitem", name = re.compile(label, re.I)).first- if item.count() > 0:+ item = wait_for_first(items, timeout_ms = 5_000)+ if item is not None:btn = itemif btn is None:soft_fail(f"nav '{label}' not found")return False+ url_before = page.url# force=True bypasses the actionability check: the post-toggle view-transition can briefly report <html> as# topmost even though the button is visible + enabled (belt-and-suspenders atop the startViewTransition# neutraliser).@@ -2000,7 +2111,16 @@ with sync_playwright() as p:except Exception as exc:soft_fail(f"nav '{label}' click failed: {exc!r}")return False- page.wait_for_timeout(800)+ if expected_url_pat:+ # Wait for the route change itself, not 800 ms: a URL that differs from the one before+ # the click, since the pattern alone may already match. The check below still decides.+ try:+ page.wait_for_url(+ lambda u: u != url_before and re.search(expected_url_pat, u) is not None,+ timeout = 5_000,+ )+ except Exception:+ passif expected_url_pat and not re.search(expected_url_pat, page.url):soft_fail(f"clicking '{label}' didn't change url to /{expected_url_pat}; current: {page.url}"@@ -2015,26 +2135,31 @@ with sync_playwright() as p:plus_btn = page.get_by_role("button", name = re.compile(r"Tools and attachments", re.I)).firstif plus_btn.count() > 0:click_forced(plus_btn)- page.wait_for_timeout(400)- compare_item = page.get_by_role("menuitem", name = re.compile(r"Compare chat", re.I)).first+ compare_items = page.get_by_role("menuitem", name = re.compile(r"Compare chat", re.I))+ # The menu renders its items together, so once any item is there Compare's absence is real.+ wait_for_first(page.get_by_role("menuitem"), timeout_ms = 5_000)+ compare_item = compare_items.firstif compare_item.count() == 0:# Compare chat moved into the "More" submenu; hover (then click as fallback) to open it.more_trigger = page.get_by_role("menuitem", name = re.compile(r"^More$", re.I)).firstif more_trigger.count() > 0:more_trigger.hover()- page.wait_for_timeout(400)- compare_item = page.get_by_role(- "menuitem", name = re.compile(r"Compare chat", re.I)- ).first- if compare_item.count() == 0:+ compare_item = wait_for_first(compare_items, timeout_ms = 5_000)+ if compare_item is None:click_forced(more_trigger)- page.wait_for_timeout(400)- compare_item = page.get_by_role(- "menuitem", name = re.compile(r"Compare chat", re.I)- ).first+ compare_item = wait_for_first(compare_items, timeout_ms = 5_000)+ compare_item = compare_items.firstif compare_item.count() > 0:+ url_before = page.urlclick_forced(compare_item)- page.wait_for_timeout(800)+ try:+ # The URL before the click may already carry a query (?new=...), so wait for a new one.+ page.wait_for_url(+ lambda u: u != url_before and re.search(r"/chat\?", u) is not None,+ timeout = 10_000,+ )+ except Exception:+ passif not re.search(r"/chat\?", page.url):soft_fail(f"'Compare chat' didn't open compare; current: {page.url}")else:@@ -2046,10 +2171,14 @@ with sync_playwright() as p:search_btn = page.get_by_role("button", name = re.compile(r"^search$", re.I)).firstif search_btn.count() > 0:search_btn.click()- page.wait_for_timeout(500)+ search_dialog = page.get_by_role("dialog")+ wait_for_first(search_dialog, timeout_ms = 10_000)shoot("13-search-dialog")page.keyboard.press("Escape")- page.wait_for_timeout(300)+ try:+ expect(search_dialog).to_have_count(0, timeout = 10_000)+ except AssertionError:+ pass # best-effort, as the fixed pause wasclick_nav("Recipes", r"/data-recipes")shoot("14-recipes")page.goto(f"{BASE}/chat")@@ -2058,11 +2187,12 @@ with sync_playwright() as p:if acct.count() > 0:step("Developer (API) tab via account menu")acct.click()- page.wait_for_timeout(400)+ wait_for_first(page.locator('[role="menu"][data-state="open"]'), timeout_ms = 5_000)dev = page.get_by_role("menuitem", name = re.compile(r"developer|api", re.I)).firstif dev.count() > 0:dev.click()- page.wait_for_timeout(800)+ # The Settings dialog it opens, instead of 800 ms; both lookups below were already best-effort.+ wait_for_first(page.get_by_role("dialog"), timeout_ms = 10_000)shoot("15-developer-tab")create_btn = page.get_by_role("button",@@ -2076,7 +2206,10 @@ with sync_playwright() as p:if keys_section.count() > 0:info(f"OK API tab text: {(keys_section.text_content() or '').strip()[:80]!r}")page.keyboard.press("Escape")- page.wait_for_timeout(300)+ try:+ expect(page.get_by_role("dialog")).to_have_count(0, timeout = 10_000)+ except AssertionError:+ pass # best-effort, as the fixed pause waselse:page.keyboard.press("Escape")@@ -2086,7 +2219,15 @@ with sync_playwright() as p:# ─────────────────────────────────────────────────────step("Recipes tab: cards render + click first card")page.goto(f"{BASE}/data-recipes")- page.wait_for_timeout(1500)+ # The route is rendered once its always-present "new recipe" control is, not after 1.5 s; the+ # list is read from IndexedDB after that, behind a "Loading recipes" placeholder.+ wait_for_first(page.locator('[data-tour="recipes-new"]'), timeout_ms = 15_000)+ try:+ expect(page.locator("main").get_by_text("Loading recipes", exact = True)).to_have_count(+ 0, timeout = 15_000+ )+ except AssertionError:+ pass # the count below reports what renderedheadings = page.locator("main h2, main h3, [data-recipe], a[href*='/data-recipes/']")n_cards = headings.count()info(f"Recipes route headings/cards: {n_cards}")@@ -2094,8 +2235,18 @@ with sync_playwright() as p:if n_cards > 0:try:headings.first.scroll_into_view_if_needed()+ list_url = page.urlheadings.first.click()- page.wait_for_timeout(1200)+ # An in-app route change: wait for the recipe's own URL, since the list page's <main>+ # would satisfy a settle check before the detail route commits.+ try:+ page.wait_for_url(+ lambda u: u != list_url and re.search(r"/data-recipes/[^/?#]+", u) is not None,+ timeout = 10_000,+ )+ except Exception:+ info(f"WARN recipe card click did not open a recipe route; current: {page.url}")+ wait_for_settled(page.locator("main"))shoot("15c-recipes-first-card")info("OK clicked first recipe card")except Exception as exc:@@ -2128,7 +2279,8 @@ with sync_playwright() as p:if attach.count() > 0:# Only hover -- clicking would block on the native file dialog.attach.hover()- page.wait_for_timeout(200)+ # The screenshot is of the hover tooltip; wait for it rather than 200 ms (best-effort).+ wait_for_first(page.get_by_role("tooltip"), timeout_ms = 5_000)shoot("16-attachment-hover")# ─────────────────────────────────────────────────────@@ -2275,7 +2427,10 @@ with sync_playwright() as p:# 17. Persisted monitor auth boundary, then shutdown. A monitor left open# must stay dormant on /login and resume after successful authentication.# ─────────────────────────────────────────────────────- step("persisted monitor stays dormant on /login and resumes after auth")+ # Its own step, with the UI budget: nothing here retries, and several of these calls+ # (page.evaluate among them) have no timeout of their own. A main run was seen to wedge+ # here and sit out the whole 720 s inactivity budget.+ step("persisted monitor: reset the browser session and open a fresh page")# Start fresh after the CLI rotation invalidates this browser session.# Stay in the SAME context: it keeps the init script and costs nothing to reuse.try:@@ -2306,6 +2461,7 @@ with sync_playwright() as p:page = _fresh_pagelogin_system_request_count = len(system_requests)+ step("persisted monitor stays dormant on /login and resumes after auth", NO_STEP_CEILING)# Re-login with NEW2 for a valid /api/shutdown token._tolerated_nav = ("ERR_ABORTED", "interrupted by another navigation")# A slow CI runner can make this re-login navigation time out even with the server healthy, so retry the whole@@ -2324,7 +2480,20 @@ with sync_playwright() as p:pw_field = page.locator("#password")pw_field.wait_for(state = "visible", timeout = 60_000)page.keyboard.press("Control+,")- page.wait_for_timeout(5_500)+ # A "nothing may happen" window, so its length stays: no /api/system request may go out+ # while /login is up, and 5.5 s spans more than one monitor poll interval. It is watched+ # rather than slept through, so a request that does go out fails the check at once+ # instead of at the end of the window.+ try:+ wait_until(+ lambda: len(system_requests) != login_system_request_count,+ timeout_s = 5.5,+ what = "an /api/system request while /login is active",+ interval_s = 0.1,+ page = page,+ )+ except TimeoutError:+ pass # the window passed with no request, which is what is assertedif len(system_requests) != login_system_request_count:raise AssertionError("persisted monitor requested /api/system while /login was active"@@ -2420,12 +2589,11 @@ with sync_playwright() as p:if acct_btn.count() == 0:fail("account menu button missing -- can't reach Shutdown")acct_btn.click()- page.wait_for_timeout(400)- shutdown_item = page.get_by_role(- "menuitem",- name = re.compile(r"^\s*Shutdown\s*$", re.I),- ).first- if shutdown_item.count() == 0:+ shutdown_item = wait_for_first(+ page.get_by_role("menuitem", name = re.compile(r"^\s*Shutdown\s*$", re.I)),+ timeout_ms = 10_000,+ )+ if shutdown_item is None:fail("Shutdown menuitem not in account menu")shutdown_item.click()shoot("19-shutdown-dialog")