Add ECW/HTTP planner scripts, browser crash recovery, and Planner error-overlay handling

- Add ecw_download.py and onedrive_planner_http.py for ECW and HTTP-based Planner access
- Add build_exe.bat for packaging
- Restart Chrome automatically on crash mid-run in planner_download.py
- Detect and recover from Planner "something went wrong" error overlays
- Improve board-load waiting, sort-descending menu matching, and screenshot timestamp rendering in chrome.py/onedrive_planner.py

src/chrome.py

@@ -397,10 +397,33 @@ def _is_image_mostly_black(path: Path, *, threshold: int = 18) -> bool:
         return False
 
 
+def _load_timestamp_font(size: int = 22):
+    """Return a PIL ImageFont at the requested size, falling back to the built-in bitmap font."""
+    try:
+        from PIL import ImageFont
+
+        candidates = [
+            "C:/Windows/Fonts/segoeui.ttf",
+            "C:/Windows/Fonts/arial.ttf",
+            "C:/Windows/Fonts/calibri.ttf",
+            "C:/Windows/Fonts/verdana.ttf",
+        ]
+        for path in candidates:
+            try:
+                return ImageFont.truetype(path, size)
+            except Exception:
+                continue
+    except Exception:
+        pass
+    try:
+        from PIL import ImageFont
+        return ImageFont.load_default()
+    except Exception:
+        return None
+
+
 def _capture_browser_with_timestamp_bar(driver: webdriver.Chrome, path: Path) -> bool:
-    """
-    Fallback when desktop capture is black: maximized browser PNG plus a taskbar-style clock strip.
-    """
+    """Browser viewport PNG with a timestamp bar showing machine time."""
     try:
         import io
 
@@ -409,18 +432,36 @@ def _capture_browser_with_timestamp_bar(driver: webdriver.Chrome, path: Path) ->
         prepare_browser_for_screenshot(driver)
         png = driver.get_screenshot_as_png()
         browser = Image.open(io.BytesIO(png)).convert("RGB")
-        bar_h = 48
-        canvas = Image.new("RGB", (browser.width, browser.height + bar_h), (30, 30, 30))
+        bar_h = 36
+        canvas = Image.new("RGB", (browser.width, browser.height + bar_h), (24, 24, 24))
         canvas.paste(browser, (0, 0))
         draw = ImageDraw.Draw(canvas)
-        stamp = datetime.now().strftime("%I:%M %p   %m/%d/%Y").lstrip("0")
-        draw.text((max(browser.width - 260, 8), browser.height + 14), stamp, fill=(255, 255, 255))
+        stamp = datetime.now().strftime("%I:%M:%S %p   %m/%d/%Y").lstrip("0")
+        font = _load_timestamp_font(22)
+        y = browser.height + (bar_h - 22) // 2
+        if font:
+            draw.text((12, y), stamp, fill=(220, 220, 220), font=font)
+        else:
+            draw.text((12, y), stamp, fill=(220, 220, 220))
         canvas.save(str(path), "PNG")
         return path.is_file() and path.stat().st_size > 0
     except Exception:
         return False
 
 
+def capture_browser_screenshot(driver: webdriver.Chrome, path: Path) -> bool:
+    """Capture only the Chromium browser viewport — no dependency on the host desktop."""
+    path = Path(path)
+    path.parent.mkdir(parents=True, exist_ok=True)
+    log_step("Browser screenshot", path=str(path))
+    if _capture_browser_with_timestamp_bar(driver, path):
+        log_step("Browser screenshot saved", path=str(path))
+        return True
+    driver.save_screenshot(str(path))
+    log_step("Viewport screenshot saved", path=str(path))
+    return path.is_file()
+
+
 def capture_fullscreen_screenshot(driver: webdriver.Chrome, path: Path) -> bool:
     """
     Full-screen capture with system taskbar visible when an interactive desktop exists.

src/onedrive_planner.py

@@ -19,6 +19,7 @@ from pathlib import Path
 
 import pyotp
 from chrome import DOWNLOADS_DIR as _CHROME_DOWNLOADS_DIR
+from chrome import capture_browser_screenshot
 from chrome import capture_fullscreen_screenshot
 from config import DEFAULT_ONEDRIVE_URL, DEFAULT_PLANNER_URL, PROJECT_ROOT, TODAY_ISO_DATE
 from debug_log import write_agent_session_debug as _agent_session_log
@@ -367,11 +368,15 @@ def _on_planner_plan_board(driver: WebDriver) -> bool:
 
 
 def _plan_load_error_present(driver: WebDriver) -> bool:
-    """True when Planner shows 'Couldn't open your plan' (URL may still be /plan/)."""
+    """True when Planner shows any error overlay that blocks normal operation."""
     phrases = (
         "couldn't open your plan",
         "could not open your plan",
         "try again in a few minutes",
+        "something went wrong",
+        "something's not right",
+        "we couldn't load",
+        "please try again",
     )
     try:
         blob = (driver.find_element(By.TAG_NAME, "body").text or "").lower()
@@ -380,6 +385,33 @@ def _plan_load_error_present(driver: WebDriver) -> bool:
         return False
 
 
+def _recover_planner_error(driver: WebDriver) -> bool:
+    """Detect and dismiss a Planner error overlay ('Something went wrong', etc).
+
+    Tries clicking the 'Try again' button first; falls back to a full page refresh.
+    Returns True if an error was detected (regardless of whether recovery succeeded).
+    """
+    if not _plan_load_error_present(driver):
+        return False
+    log_step("Planner error overlay detected — attempting recovery")
+    if _click_plan_try_again(driver):
+        time.sleep(4)
+        if not _plan_load_error_present(driver):
+            log_step("Planner error overlay cleared via Try again")
+            return True
+    # Fallback: full page refresh
+    try:
+        log_step("Refreshing page to clear Planner error overlay")
+        driver.refresh()
+        time.sleep(6)
+        driver.switch_to.default_content()
+        _switch_to_planner_context(driver)
+        log_step("Page refreshed — error overlay cleared")
+    except Exception as exc:
+        log_step("Refresh failed during error recovery", error=str(exc)[:120])
+    return True
+
+
 def _plan_page_ready(driver: WebDriver) -> bool:
     """Plan UI is usable — not the error page."""
     if _plan_load_error_present(driver):
@@ -1157,6 +1189,7 @@ def open_my_plans_shared(driver: WebDriver) -> None:
 
     This matches your current UI flow and avoids search-based plan opening.
     """
+    _recover_planner_error(driver)
     log_step("Navigate: My plans -> Shared")
     log_url(driver, "before")
     driver.switch_to.default_content()
@@ -1374,7 +1407,26 @@ def _open_facility_from_shared_list(driver: WebDriver, facility_name: str) -> No
             _type_facility_filter(driver, term)
             if _filter_shows_no_results(driver):
                 log_step("Filter returned no results", query=term[:60], facility=facility_name)
-                continue
+                # A Planner error overlay may be blocking the search UI — try to recover
+                if _recover_planner_error(driver):
+                    log_step("Recovered from error overlay during filter search", facility=facility_name)
+                    # Re-open My plans -> Shared after the page refreshed
+                    _click_any(driver, _MY_PLANS_NAV_LOCATORS, timeout=WAIT_LONG)
+                    time.sleep(2)
+                    shared_locators = [
+                        (By.XPATH, "//*[self::button or self::div or self::span][normalize-space()='Shared']"),
+                        (By.XPATH, "//*[@role='tab' and normalize-space()='Shared']"),
+                    ]
+                    try:
+                        _click_any(driver, shared_locators, timeout=WAIT_LONG)
+                        time.sleep(2)
+                    except TimeoutException:
+                        pass
+                    _type_facility_filter(driver, term)
+                    if _filter_shows_no_results(driver):
+                        continue
+                else:
+                    continue
             rows = _shared_list_data_rows(driver)
             log_step(
                 "Filter results",
@@ -1611,6 +1663,53 @@ def _ensure_board_view(driver: WebDriver) -> None:
         log_step("Board tab not found (continuing on current view)")
 
 
+def _wait_for_board_loaded(driver: WebDriver, timeout: int = 20) -> None:
+    """Switch to Board tab and wait until board content is visible before screenshotting."""
+    board_locators = [
+        (By.XPATH, "//*[@role='tab' and normalize-space()='Board']"),
+        (By.XPATH, "//*[self::button or self::div][normalize-space()='Board']"),
+    ]
+
+    # Click the Board tab (ignore if already on board or tab not found)
+    try:
+        tab = _first_clickable(driver, board_locators, timeout=8)
+        # Only click if the tab isn't already selected
+        selected = (tab.get_attribute("aria-selected") or "").lower()
+        if selected != "true":
+            log_step("Switch to Board view")
+            _js_click(driver, tab)
+        else:
+            log_step("Board tab already active")
+    except TimeoutException:
+        log_step("Board tab not found (continuing on current view)")
+
+    # Poll until board content is visible (task cards, bucket columns, or Add task)
+    content_xpaths = [
+        "//button[@data-testid='attachment-indicator']",
+        "//*[@role='checkbox']",
+        "//label[@role='checkbox']",
+        "//*[contains(@data-testid,'TaskCard') or contains(@class,'TaskCard')]",
+        "//*[@role='list' and ancestor::*[@role='group']]",
+        "//*[normalize-space()='Add task' and ancestor::*[@role='group']]",
+        "//*[normalize-space()='Add task']",
+    ]
+
+    log_step("Waiting for board content to load")
+    deadline = time.time() + timeout
+    while time.time() < deadline:
+        for xpath in content_xpaths:
+            try:
+                for el in driver.find_elements(By.XPATH, xpath)[:6]:
+                    if el.is_displayed():
+                        log_step("Board content visible — ready for screenshot")
+                        return
+            except Exception:
+                continue
+        time.sleep(0.5)
+
+    log_step("Board content wait timed out — proceeding with screenshot anyway")
+
+
 def _ensure_grid_view(driver: WebDriver) -> None:
     grid_locators = [
         (By.CSS_SELECTOR, "button#gridViewPivot"),
@@ -1642,13 +1741,13 @@ def _is_status_sorted_descending(header) -> bool:
 
 
 def _sort_descending_menu_selected(driver: WebDriver) -> bool:
-    """True when the column header menu shows Sort Descending as active."""
+    """True when the column header menu shows Sort descending as active."""
     probes = [
-        "//button[@name='Sort Descending' and @aria-checked='true']",
-        "//button[@name='Sort Descending' and contains(@class,'is-checked')]",
-        "//button[@name='Sort Descending']//*[@data-icon-name='CheckMark']",
+        "//button[@name='Sort descending' and @aria-checked='true']",
+        "//button[@name='Sort descending' and contains(@class,'is-checked')]",
+        "//button[@name='Sort descending']//*[@data-icon-name='CheckMark']",
         (
-            "//button[@name='Sort Descending']"
+            "//button[@name='Sort descending']"
             "/ancestor::li[contains(@class,'is-checked') or contains(@class,'is-selected')]"
         ),
     ]
@@ -1692,13 +1791,18 @@ def _ensure_status_sort_descending(driver: WebDriver) -> None:
         return
 
     sort_locators = [
+        # Exact match — actual Planner DOM uses lowercase 'd'
+        (By.XPATH, "//button[@name='Sort descending' and @role='menuitem']"),
+        (By.XPATH, "//ul[contains(@class,'ContextualMenu')]//button[normalize-space()='Sort descending']"),
+        (By.XPATH, "//button[@role='menuitem' and normalize-space()='Sort descending']"),
+        (By.XPATH, "//span[normalize-space()='Sort descending']/ancestor::button[@role='menuitem']"),
+        # Uppercase fallback (in case Planner localises differently)
         (By.XPATH, "//button[@name='Sort Descending' and @role='menuitem']"),
-        (
-            By.XPATH,
-            "//ul[contains(@class,'ContextualMenu')]"
-            "//button[normalize-space()='Sort Descending']",
-        ),
         (By.XPATH, "//button[@role='menuitem' and normalize-space()='Sort Descending']"),
+        # Broad contains — catches any capitalisation
+        (By.XPATH, "//*[@role='menuitem' and contains(normalize-space(),'Sort descending')]"),
+        (By.XPATH, "//*[@role='menuitem' and contains(normalize-space(),'Sort Descending')]"),
+        (By.XPATH, "//li[contains(@class,'ms-ContextualMenu-item')]//button[contains(normalize-space(),'escending')]"),
     ]
     try:
         _click_any(driver, sort_locators, timeout=5)
@@ -1712,8 +1816,30 @@ def _ensure_status_sort_descending(driver: WebDriver) -> None:
         except TimeoutException:
             log_step("Status column sort descending clicked")
     except TimeoutException:
-        log_step("Sort Descending menu item not found (continuing on current sort)")
-        _dismiss_open_menu(driver)
+        log_step("Sort descending menu item not found via XPath — trying JS click")
+        clicked = False
+        try:
+            clicked = driver.execute_script(
+                """
+                var items = Array.from(document.querySelectorAll(
+                    '[role="menuitem"], [role="menuitem"] button, button[role="menuitem"]'
+                ));
+                var target = items.find(function(el) {
+                    var t = (el.innerText || el.textContent || '').trim().toLowerCase();
+                    return t === 'sort descending';
+                });
+                if (target) { target.click(); return true; }
+                return false;
+                """
+            )
+        except Exception:
+            pass
+        if clicked:
+            time.sleep(1)
+            log_step("Sort descending clicked via JS")
+        else:
+            log_step("Sort descending not found (continuing on current sort)")
+            _dismiss_open_menu(driver)
 
 
 def _grid_probes_present(driver: WebDriver) -> bool:
@@ -1902,11 +2028,14 @@ def _scroll_grid_down(driver: WebDriver, pixels: int = 450) -> None:
     time.sleep(0.4)
 
 
-def _scan_grid_not_started_titles(driver: WebDriver, *, max_scrolls: int = 100) -> list[str]:
-    """Scroll the virtualized grid and collect Not started task titles."""
+def _scan_grid_not_started_titles(driver: WebDriver, *, max_scrolls: int = 100) -> list[tuple[str, int]]:
+    """Scroll the virtualized grid and collect (title, aria-rowindex) for Not started tasks.
+
+    Uses aria-rowindex as the unique key so duplicate task names are all captured.
+    """
     _scroll_grid_to_top(driver)
-    seen: set[str] = set()
-    titles: list[str] = []
+    seen_indices: set[int] = set()
+    tasks: list[tuple[str, int]] = []
     stagnant = 0
     for _ in range(max_scrolls):
         new_this_pass = 0
@@ -1914,14 +2043,15 @@ def _scan_grid_not_started_titles(driver: WebDriver, *, max_scrolls: int = 100)
             status = _grid_row_status(row)
             if status != "not started":
                 continue
+            idx_raw = row.get_attribute("aria-rowindex") or ""
+            row_idx = int(idx_raw) if idx_raw.isdigit() else -1
+            if row_idx < 1 or row_idx in seen_indices:
+                continue
             title = _grid_row_task_name(row).strip()
             if not title or _is_weak_task_title_key(title.lower()):
                 continue
-            key = title.lower()
-            if key in seen:
-                continue
-            seen.add(key)
-            titles.append(title)
+            seen_indices.add(row_idx)
+            tasks.append((title, row_idx))
             new_this_pass += 1
         if new_this_pass == 0:
             stagnant += 1
@@ -1931,8 +2061,22 @@ def _scan_grid_not_started_titles(driver: WebDriver, *, max_scrolls: int = 100)
             stagnant = 0
         _scroll_grid_down(driver)
     _scroll_grid_to_top(driver)
-    log_step("Grid not-started tasks scanned", count=len(titles))
-    return titles
+    log_step("Grid not-started tasks scanned", count=len(tasks))
+    return tasks
+
+
+def _find_grid_row_by_index(driver: WebDriver, row_index: int, *, max_scrolls: int = 100):
+    """Find a grid row by its aria-rowindex — stable even when multiple rows share the same title."""
+    if row_index < 1:
+        return None
+    _scroll_grid_to_top(driver)
+    for _ in range(max_scrolls):
+        for row in _grid_visible_data_rows(driver):
+            idx_raw = row.get_attribute("aria-rowindex") or ""
+            if idx_raw.isdigit() and int(idx_raw) == row_index:
+                return row
+        _scroll_grid_down(driver)
+    return None
 
 
 def _find_grid_row_by_title(driver: WebDriver, title: str, *, max_scrolls: int = 100):
@@ -6047,13 +6191,13 @@ def save_board_screenshot(
         safe = _sanitize_folder_name(facility_name)
         taken_at = datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
         path = shots_dir / f"{safe}_{taken_at}.png"
-        capture_fullscreen_screenshot(driver, path)
+        capture_browser_screenshot(driver, path)
         log_step(
             "Board screenshot saved",
             facility=facility_name,
             tasks=task_count,
             path=str(path),
-            fullscreen="true",
+            fullscreen="false",
             view="board",
         )
         if task_count:
@@ -6114,6 +6258,7 @@ def process_facility(
     try:
         open_my_plans_shared(driver)
         _open_facility_from_shared_list(driver, facility_name)
+        _wait_for_board_loaded(driver)
         save_board_screenshot(driver, facility_name)
         _ensure_grid_view(driver)
         _ensure_status_sort_descending(driver)
@@ -6135,8 +6280,8 @@ def process_facility(
         log_step("Grid not visible — waiting for grid to settle", facility=facility_name)
         time.sleep(6)
 
-    task_titles = _scan_grid_not_started_titles(driver)
-    task_count = len(task_titles)
+    task_entries = _scan_grid_not_started_titles(driver)
+    task_count = len(task_entries)
     log_step(
         "Scan tasks on grid",
         task_count=task_count,
@@ -6147,19 +6292,21 @@ def process_facility(
         return FacilityRunResult()
 
     download_dir_ready = False
-    processed_titles: set[str] = set()
+    processed_indices: set[int] = set()
 
-    for task_num, grid_title in enumerate(task_titles, start=1):
+    for task_num, (grid_title, row_index) in enumerate(task_entries, start=1):
         title_key = grid_title.strip().lower()
-        if title_key in processed_titles:
+        if row_index in processed_indices:
             continue
         _close_task_panel(driver)
         task_title = ""
         try:
-            row = _find_grid_row_by_title(driver, grid_title)
+            # Locate by row index first (handles duplicate task names correctly),
+            # fall back to title search if the row has no aria-rowindex.
+            row = _find_grid_row_by_index(driver, row_index) or _find_grid_row_by_title(driver, grid_title)
             if row is None:
                 log_step("Grid row not found", title=grid_title[:60], facility=facility_name)
-                processed_titles.add(title_key)
+                processed_indices.add(row_index)
                 continue
 
             row_status = _grid_row_status(row)
@@ -6170,7 +6317,7 @@ def process_facility(
                     status=row_status or "",
                     facility=facility_name,
                 )
-                processed_titles.add(title_key)
+                processed_indices.add(row_index)
                 continue
 
             log_step(
@@ -6180,7 +6327,7 @@ def process_facility(
                 facility=facility_name,
             )
             # Re-find row right before click — virtualized grid can shift rows.
-            row = _find_grid_row_by_title(driver, grid_title)
+            row = _find_grid_row_by_index(driver, row_index) or _find_grid_row_by_title(driver, grid_title)
             if row is None or not _open_grid_row_details(driver, row):
                 log_step("Grid task open failed", title=grid_title[:60], facility=facility_name)
                 continue
@@ -6188,7 +6335,7 @@ def process_facility(
             # Grid Status column already filtered to Not started — go straight
             # to Attachments tab (no panel status gate).
             task_title = _clean_grid_task_title(grid_title)
-            processed_titles.add(task_title.strip().lower() or title_key)
+            processed_indices.add(row_index)
             log_step("Task opened from grid", title=task_title[:60])
             time.sleep(PANEL_TAB_LOAD_DELAY)
 
@@ -6280,10 +6427,7 @@ def process_facility(
             )
             _log.warning("Task %d failed: %s", task_num, exc)
             _close_task_panel(driver)
-            if task_title:
-                processed_titles.add(task_title.strip().lower())
-            elif title_key:
-                processed_titles.add(title_key)
+            processed_indices.add(row_index)
 
     _remove_empty_directory(dest_folder)
     log_step(

src/planner_download.py

@@ -58,6 +58,64 @@ from debug_log import write_debug as _agent_log
 _log = logging.getLogger(__name__)
 
 
+def _is_browser_dead(exc: Exception) -> bool:
+    """True when Chrome itself has crashed and the WebDriver session is gone."""
+    err = str(exc).lower()
+    return any(phrase in err for phrase in (
+        "invalid session id",
+        "session deleted",
+        "browser has closed the connection",
+        "chrome not reachable",
+        "cannot connect to chrome",
+        "no such window",
+    ))
+
+
+def _restart_browser(cfg: "AppConfig") -> "webdriver.Chrome | None":
+    """Kill Chrome, relaunch it, and re-login to Planner. Returns new driver on success."""
+    log_step("Browser crash detected — restarting Chrome")
+    _log.warning("Restarting Chrome after crash...")
+    try:
+        stop_chrome_browser()
+    except Exception:
+        pass
+    time.sleep(4)
+    try:
+        if not start_chrome_debug(
+            cfg.proxy_ip, cfg.proxy_port, cfg.proxy_username, cfg.proxy_password
+        ):
+            log_step("Chrome restart failed — start_chrome_debug returned False")
+            return None
+        new_driver = attach_browser(create_download_dir=False)
+        if new_driver is None:
+            log_step("Chrome restart failed — attach_browser returned None")
+            return None
+        consolidate_browser_to_single_tab(new_driver)
+        if cfg.proxy_username and cfg.proxy_password:
+            dismiss_proxy_auth_popup(
+                new_driver,
+                cfg.proxy_username.strip(),
+                str(cfg.proxy_password),
+                wait_secs=8.0,
+                max_attempts=3,
+            )
+            time.sleep(2)
+        login_and_open_planner(
+            new_driver,
+            cfg.onedrive_username,
+            cfg.onedrive_password,
+            cfg.microsoft_totp_secret,
+            planner_url=cfg.planner_url,
+        )
+        log_step("Chrome restarted and Planner re-opened after crash")
+        _log.info("Chrome restarted successfully after crash.")
+        return new_driver
+    except Exception as exc:
+        log_step("Browser restart failed", error=str(exc)[:200])
+        _log.error("Browser restart failed: %s", exc)
+        return None
+
+
 def _show_error_dialog(title: str, message: str) -> None:
     try:
         root = tk.Tk()
@@ -624,7 +682,10 @@ def run_download_workflow(
                 log_step("Facility failed", facility=fac, error=err_msg[:200], status=status)
                 _log.error("%s: %s", fac, err_msg)
                 if screenshot_on_error and driver is not None:
-                    save_failure_screenshot(driver, fac, failure_log_dir=failure_log_dir)
+                    try:
+                        save_failure_screenshot(driver, fac, failure_log_dir=failure_log_dir)
+                    except Exception:
+                        pass
                 failures.append((fac, err_msg, status))
                 results[fac] = {
                     "original": [],
@@ -644,6 +705,21 @@ def run_download_workflow(
                     completed_facilities=sorted(completed),
                 )
 
+                # Chrome crashed — restart it so remaining facilities can still run
+                if _is_browser_dead(exc):
+                    _log.warning(
+                        "Browser crash detected after '%s' — restarting Chrome to continue run",
+                        fac,
+                    )
+                    new_driver = _restart_browser(cfg)
+                    if new_driver is not None:
+                        driver = new_driver
+                    else:
+                        _log.error(
+                            "Could not restart browser after crash — aborting remaining facilities"
+                        )
+                        break
+
         summary_path = (
             PROJECT_ROOT / "Output" / TODAY_ISO_DATE / "Summary"
             / f"Download_Summary_{TODAY_ISO_DATE}.xlsx"