feat: add adaptive GPU scheduling

tests/test_modelscope_discovery.py

@@ -11,6 +11,7 @@ if str(PACKAGE_DIR) in sys.path:
 sys.path.insert(0, str(PACKAGE_DIR))
 
 from hf_discovery import HuggingFaceDiscovery  # noqa: E402
+from http_json import HttpJsonError  # noqa: E402
 
 
 class RecordingHttpClient:
@@ -22,6 +23,30 @@ class RecordingHttpClient:
         return {"success": True, "data": {"models": []}}
 
 
+class RateLimitedPageClient:
+    def __init__(self) -> None:
+        self.page_calls: list[int] = []
+        self.page_two_attempts = 0
+
+    def request_json(self, _method: str, _path: str, *, query: dict) -> dict:
+        page = int(query["page_number"])
+        self.page_calls.append(page)
+        if page == 2:
+            self.page_two_attempts += 1
+            if self.page_two_attempts == 1:
+                raise HttpJsonError("rate limited", status_code=429)
+        start = (page - 1) * 50
+        items = [
+            {
+                "id": f"owner/model-{index}",
+                "downloads": 100,
+                "last_modified": "2026-01-01T00:00:00Z",
+            }
+            for index in range(start, start + 50)
+        ]
+        return {"success": True, "data": {"models": items}}
+
+
 class ModelScopeDiscoveryTests(unittest.TestCase):
     def test_openapi_page_size_never_exceeds_platform_limit(self) -> None:
         http_client = RecordingHttpClient()
@@ -36,6 +61,29 @@ class ModelScopeDiscoveryTests(unittest.TestCase):
         self.assertEqual([], models)
         self.assertEqual(50, http_client.queries[0]["page_size"])
 
+    def test_rate_limited_page_keeps_previous_pages_and_resumes_from_cache(self) -> None:
+        http_client = RateLimitedPageClient()
+        discovery = HuggingFaceDiscovery(
+            http_client=http_client,  # type: ignore[arg-type]
+            page_interval_seconds=0,
+            page_cache_ttl_seconds=60,
+        )
+
+        partial = discovery.list_recent_models(
+            pipeline_tags=["text-generation"],
+            limit=100,
+            min_downloads=0,
+        )
+        resumed = discovery.list_recent_models(
+            pipeline_tags=["text-generation"],
+            limit=100,
+            min_downloads=0,
+        )
+
+        self.assertEqual(50, len(partial))
+        self.assertEqual(100, len(resumed))
+        self.assertEqual([1, 2, 2], http_client.page_calls)
+
 
 if __name__ == "__main__":
     unittest.main()