perf(check-model-exists): eliminate SQLite connection-per-query overhead and skip redundant history checks

Root cause: 231 concurrent /check-model-exists requests on 175K-lora library
caused ~9.4s wall clock time. The bottleneck was two-fold:

1. DownloadedVersionHistoryService opened a new sqlite3.connect() for every
   query under asyncio.Lock. With a large WAL from 175K entries, each
   connect() took ~8ms. Serialized by the lock across 231 requests, the
   230th request waited ~1848ms just for lock acquisition.

2. check_model_exists always queried download history even when the model
   was found locally. The history result (hasBeenDownloaded /
   downloadedVersionIds) is only used by the UI when the model is NOT
   found locally; when found, the 'in library' indicator takes priority.

Changes:
- downloaded_version_history_service.py: added persistent _get_conn() that
  creates the SQLite connection once and reuses it across all queries
- misc_handlers.py: early-return from check_model_exists when the model
  exists locally, bypassing the history service entirely (lock skipped)

Expected: per-request wait time drops from ~1912ms to <3ms, wall clock
from ~9.4s to <0.3s for the 175K-lora user's 231-card page.

py/routes/handlers/misc_handlers.py

@@ -1791,15 +1791,19 @@ class ModelLibraryHandler:
                     exists = True
                     model_type = "embedding"
 
+                if exists:
+                    return web.json_response(
+                        {
+                            "success": True,
+                            "exists": True,
+                            "modelType": model_type,
+                            "hasBeenDownloaded": False,
+                        }
+                    )
+
                 history_service = await self._get_download_history_service()
                 has_been_downloaded = False
-                history_type = model_type
+                history_type = None
-                if history_type:
-                    has_been_downloaded = await history_service.has_been_downloaded(
-                        history_type,
-                        model_version_id,
-                    )
-                else:
                 for candidate_type in ("lora", "checkpoint", "embedding"):
                     if await history_service.has_been_downloaded(
                         candidate_type,
@@ -1812,8 +1816,8 @@ class ModelLibraryHandler:
                 return web.json_response(
                     {
                         "success": True,
-                        "exists": exists,
+                        "exists": False,
-                        "modelType": model_type if exists else history_type,
+                        "modelType": history_type,
                         "hasBeenDownloaded": has_been_downloaded,
                     }
                 )
@@ -1833,29 +1837,35 @@ class ModelLibraryHandler:
             model_type = None
             versions = []
             downloaded_version_ids = []
-            history_service = await self._get_download_history_service()
             if lora_versions:
-                model_type = "lora"
+                return web.json_response(
-                versions = self._with_downloaded_flag(lora_versions)
+                    {
-                downloaded_version_ids = await history_service.get_downloaded_version_ids(
+                        "success": True,
-                    model_type,
+                        "modelType": "lora",
-                    model_id,
+                        "versions": self._with_downloaded_flag(lora_versions),
+                        "downloadedVersionIds": [],
+                    }
                 )
-            elif checkpoint_versions:
+            if checkpoint_versions:
-                model_type = "checkpoint"
+                return web.json_response(
-                versions = self._with_downloaded_flag(checkpoint_versions)
+                    {
-                downloaded_version_ids = await history_service.get_downloaded_version_ids(
+                        "success": True,
-                    model_type,
+                        "modelType": "checkpoint",
-                    model_id,
+                        "versions": self._with_downloaded_flag(checkpoint_versions),
+                        "downloadedVersionIds": [],
+                    }
                 )
-            elif embedding_versions:
+            if embedding_versions:
-                model_type = "embedding"
+                return web.json_response(
-                versions = self._with_downloaded_flag(embedding_versions)
+                    {
-                downloaded_version_ids = await history_service.get_downloaded_version_ids(
+                        "success": True,
-                    model_type,
+                        "modelType": "embedding",
-                    model_id,
+                        "versions": self._with_downloaded_flag(embedding_versions),
+                        "downloadedVersionIds": [],
+                    }
                 )
-            else:
+
+            history_service = await self._get_download_history_service()
             for candidate_type in ("lora", "checkpoint", "embedding"):
                 candidate_downloaded_version_ids = (
                     await history_service.get_downloaded_version_ids(

py/services/downloaded_version_history_service.py

@@ -64,6 +64,7 @@ class DownloadedVersionHistoryService:
         self._db_path = db_path or _resolve_database_path()
         self._settings = settings_manager or get_settings_manager()
         self._lock = asyncio.Lock()
+        self._conn: sqlite3.Connection | None = None
         self._schema_initialized = False
         self._ensure_directory()
         self._initialize_schema()
@@ -78,6 +79,12 @@ class DownloadedVersionHistoryService:
         conn.row_factory = sqlite3.Row
         return conn
 
+    def _get_conn(self) -> sqlite3.Connection:
+        if self._conn is None:
+            self._conn = sqlite3.connect(self._db_path, check_same_thread=False)
+            self._conn.row_factory = sqlite3.Row
+        return self._conn
+
     def _initialize_schema(self) -> None:
         if self._schema_initialized:
             return
@@ -116,7 +123,7 @@ class DownloadedVersionHistoryService:
         timestamp = time.time()
 
         async with self._lock:
-            with self._connect() as conn:
+            conn = self._get_conn()
             conn.execute(
                 """
                 INSERT INTO downloaded_model_versions (
@@ -180,7 +187,7 @@ class DownloadedVersionHistoryService:
             return
 
         async with self._lock:
-            with self._connect() as conn:
+            conn = self._get_conn()
             conn.executemany(
                 """
                 INSERT INTO downloaded_model_versions (
@@ -208,7 +215,7 @@ class DownloadedVersionHistoryService:
         timestamp = time.time()
 
         async with self._lock:
-            with self._connect() as conn:
+            conn = self._get_conn()
             conn.execute(
                 """
                 INSERT INTO downloaded_model_versions (
@@ -238,7 +245,7 @@ class DownloadedVersionHistoryService:
             return False
 
         async with self._lock:
-            with self._connect() as conn:
+            conn = self._get_conn()
             row = conn.execute(
                 """
                 SELECT is_deleted_override
@@ -258,7 +265,7 @@ class DownloadedVersionHistoryService:
             return []
 
         async with self._lock:
-            with self._connect() as conn:
+            conn = self._get_conn()
             rows = conn.execute(
                 """
                 SELECT version_id
@@ -291,7 +298,7 @@ class DownloadedVersionHistoryService:
         params: list[object] = [normalized_type, *normalized_model_ids]
 
         async with self._lock:
-            with self._connect() as conn:
+            conn = self._get_conn()
             rows = conn.execute(
                 f"""
                 SELECT model_id, version_id