perf(library): optimize paginated fetch with bounded concurrent scheduling

auditui/library/client_fetch.py

@@ -3,7 +3,6 @@
 from __future__ import annotations
 
 from concurrent.futures import ThreadPoolExecutor, as_completed
-
 from typing import Any
 
 from ..types import LibraryItem, StatusCallback
@@ -76,39 +75,73 @@ class LibraryClientFetchMixin:
         if len(first_page_items) < page_size:
             return all_items
 
+        estimated_pages = self._estimate_total_pages(
+            library_response, page_size)
+        page_results = self._fetch_remaining_pages(
+            response_groups=response_groups,
+            page_size=page_size,
+            estimated_pages=estimated_pages,
+            initial_total=len(first_page_items),
+            on_progress=on_progress,
+        )
+
+        for page_num in sorted(page_results.keys()):
+            all_items.extend(page_results[page_num])
+
+        return all_items
+
+    def _estimate_total_pages(self, library_response: dict, page_size: int) -> int:
+        """Estimate total pages from API metadata with a conservative cap."""
         total_items_estimate = library_response.get(
             "total_results"
         ) or library_response.get("total")
-        if total_items_estimate:
+        if not total_items_estimate:
+            return 500
         estimated_pages = (total_items_estimate + page_size - 1) // page_size
-            estimated_pages = min(estimated_pages, 1000)
-        else:
-            estimated_pages = 500
+        return min(estimated_pages, 1000)
 
-        max_workers = 50
+    def _fetch_remaining_pages(
+        self,
+        response_groups: str,
+        page_size: int,
+        estimated_pages: int,
+        initial_total: int,
+        on_progress: StatusCallback | None = None,
+    ) -> dict[int, list[LibraryItem]]:
+        """Fetch pages 2..N with bounded in-flight requests for faster startup."""
         page_results: dict[int, list[LibraryItem]] = {}
+        max_workers = min(16, max(1, estimated_pages - 1))
+        next_page_to_submit = 2
+        stop_page = estimated_pages + 1
+        completed_count = 0
+        total_items = initial_total
 
         with ThreadPoolExecutor(max_workers=max_workers) as executor:
             future_to_page: dict = {}
 
-            for page in range(2, estimated_pages + 1):
+            while (
+                next_page_to_submit <= estimated_pages
+                and next_page_to_submit < stop_page
+                and len(future_to_page) < max_workers
+            ):
                 future = executor.submit(
-                    self._fetch_page, page, page_size, response_groups
+                    self._fetch_page,
+                    next_page_to_submit,
+                    page_size,
+                    response_groups,
                 )
-                future_to_page[future] = page
+                future_to_page[future] = next_page_to_submit
+                next_page_to_submit += 1
 
-            completed_count = 0
-            total_items = len(first_page_items)
-
-            for future in as_completed(future_to_page):
+            while future_to_page:
+                future = next(as_completed(future_to_page))
+                page_num = future_to_page.pop(future)
                 try:
                     fetched_page, items = future.result()
-                    future_to_page.pop(future, None)
-                    if not items or len(items) < page_size:
-                        for remaining_future in list(future_to_page.keys()):
-                            remaining_future.cancel()
-                        break
+                except Exception:
+                    continue
 
+                if items:
                     page_results[fetched_page] = items
                     total_items += len(items)
                     completed_count += 1
@@ -116,12 +149,21 @@ class LibraryClientFetchMixin:
                         on_progress(
                             f"Fetched {completed_count} pages ({total_items} items)..."
                         )
+                if len(items) < page_size:
+                    stop_page = min(stop_page, fetched_page)
 
-                except Exception:
-                    future_to_page.pop(future, None)
-                    pass
+                while (
+                    next_page_to_submit <= estimated_pages
+                    and next_page_to_submit < stop_page
+                    and len(future_to_page) < max_workers
+                ):
+                    next_future = executor.submit(
+                        self._fetch_page,
+                        next_page_to_submit,
+                        page_size,
+                        response_groups,
+                    )
+                    future_to_page[next_future] = next_page_to_submit
+                    next_page_to_submit += 1
 
-        for page_num in sorted(page_results.keys()):
-            all_items.extend(page_results[page_num])
-
-        return all_items
+        return page_results