Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

Tests: keep the RAG query-packing test on vec0 while the in-memory cache builds in the background (#16013)

Daniel Han86be80531b82
diff --git a/studio/backend/tests/test_rag_store.py b/studio/backend/tests/test_rag_store.py
index bd65adf1e07..94770103c18 100644
--- a/studio/backend/tests/test_rag_store.py
+++ b/studio/backend/tests/test_rag_store.py
@@ -597,6 +597,9 @@ def test_a_legacy_archive_still_gets_two_different_ends(rag_home, rag_conn):
def test_dense_packs_float32_rows_without_struct_and_the_query_once(rag_conn, monkeypatch):
np = pytest.importorskip("numpy")
+ # vec0 answers both searches: the first one queues the in-memory cache build (#15961) on a background
+ # thread, and a cache that lands between them scores in numpy, a few ulps away from vec0.
+ monkeypatch.setenv("RAG_VEC_CACHE", "0")
texts = ["alpha alpha", "hotel golf", "alpha bravo"]
for i, scope in enumerate(("kb_a", "kb_b", "kb_c")):
doc = f"d{i}"