Merge branch 'feature/rag' of github.com:unslothai/unsloth into feature/rag

studio/backend/core/rag/db.py

@@ -73,9 +73,7 @@ def _backfill_fts(conn: sqlite3.Connection) -> None:
     """One-time: seed FTS from existing vectors for rag.db files created before
     FTS5 replaced the on-disk bm25s index. Runs only when FTS is empty but
     vectors exist; idempotent thereafter."""
-    fts_seeded = conn.execute(
-        "SELECT 1 FROM rag_chunks_fts LIMIT 1"
-    ).fetchone()
+    fts_seeded = conn.execute("SELECT 1 FROM rag_chunks_fts LIMIT 1").fetchone()
     if fts_seeded is not None:
         return
     has_vectors = conn.execute("SELECT 1 FROM rag_vectors LIMIT 1").fetchone()

studio/backend/core/rag/vector_store.py

@@ -89,9 +89,7 @@ def upsert_chunks(scope: str, points: Iterable[dict]) -> None:
         rows,
     )
     if fts_rows:
-        conn.executemany(
-            "DELETE FROM rag_chunks_fts WHERE chunk_id = ?", fts_delete
-        )
+        conn.executemany("DELETE FROM rag_chunks_fts WHERE chunk_id = ?", fts_delete)
         conn.executemany(
             "INSERT INTO rag_chunks_fts (text, chunk_id, scope) VALUES (?, ?, ?)",
             fts_rows,

studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts

@@ -255,7 +255,8 @@ export function useChatModelRuntime() {
     [],
   );
 
-  const refresh = useCallback(async () => {
+  const refresh = useCallback(async (options?: { signal?: AbortSignal }) => {
+    const signal = options?.signal;
     setModelsError(null);
     try {
       const [listRes, statusRes, lorasRes] = await Promise.all([
@@ -264,6 +265,11 @@ export function useChatModelRuntime() {
         listLoras(),
       ]);
 
+      // Cancellation can land while the requests above are in flight (e.g. the
+      // user cancels a load during this refresh). Bail before writing any
+      // backend state back into the store -- cancelLoading already cleared it.
+      if (signal?.aborted) return;
+
       setModels(listRes.models.map(toChatModelSummary));
       setLoras(lorasRes.loras.map(toLoraSummary));
 
@@ -405,6 +411,7 @@ export function useChatModelRuntime() {
         });
       }
     } catch (error) {
+      if (signal?.aborted) return;
       const message =
         error instanceof Error ? error.message : "Failed to load models";
       setModelsError(message);
@@ -766,7 +773,7 @@ export function useChatModelRuntime() {
                 store.setParams({ ...store.params, ...p });
               }
             }
-            await refresh();
+            await refresh({ signal: abortCtrl.signal });
           } catch (error) {
             // Skip rollback if user cancelled -- model is already being unloaded.
             if (abortCtrl.signal.aborted) throw error;
@@ -1061,6 +1068,8 @@ export function useChatModelRuntime() {
 
         try {
           await performLoad();
+          // User cancelled mid-refresh; cancelLoading handles teardown.
+          if (abortCtrl.signal.aborted) return;
           if (loadToastDismissedRef.current) {
             toast.success(`${toastDisplayName} loaded`, {
               classNames: MODEL_LOADED_TOAST_CLASSNAMES,

tests/python/test_rag_bm25.py

@@ -46,7 +46,9 @@ def test_lexical_search_roundtrip(isolated_rag_db):
         scope,
         [
             _chunk("c1", "the quick brown fox jumps over the lazy dog", index = 0),
-            _chunk("c2", "machine learning models predict outputs from inputs", index = 1),
+            _chunk(
+                "c2", "machine learning models predict outputs from inputs", index = 1
+            ),
             _chunk("c3", "fox terriers are small dogs", index = 2),
         ],
     )