new: add local_files_only propagation to fastembed mixin

qdrant_client/async_qdrant_fastembed.py

@@ -63,6 +63,7 @@ class AsyncQdrantFastembedMixin(AsyncQdrantBase):
     def __init__(self, **kwargs: Any):
         self._embedding_model_name: Optional[str] = None
         self._sparse_embedding_model_name: Optional[str] = None
+        self.local_files_only = kwargs.get("local_files_only", False)
         try:
             from fastembed import SparseTextEmbedding, TextEmbedding
 
@@ -117,11 +118,13 @@ def set_model(
                 DeprecationWarning,
                 stacklevel=2,
             )
+        local_files_only = kwargs.pop("local_files_only", self.local_files_only)
         self._get_or_init_model(
             model_name=embedding_model_name,
             cache_dir=cache_dir,
             threads=threads,
             providers=providers,
+            local_files_only=local_files_only,
             **kwargs,
         )
         self._embedding_model_name = embedding_model_name
@@ -132,6 +135,7 @@ def set_sparse_model(
         cache_dir: Optional[str] = None,
         threads: Optional[int] = None,
         providers: Optional[Sequence["OnnxProvider"]] = None,
+        **kwargs: Any,
     ) -> None:
         """
         Set sparse embedding model to use for hybrid search over documents in combination with dense embeddings.
@@ -152,12 +156,15 @@ def set_sparse_model(
         Returns:
             None
         """
+        local_files_only = kwargs.pop("local_files_only", self.local_files_only)
         if embedding_model_name is not None:
             self._get_or_init_sparse_model(
                 model_name=embedding_model_name,
                 cache_dir=cache_dir,
                 threads=threads,
                 providers=providers,
+                local_files_only=local_files_only,
+                **kwargs,
             )
         self._sparse_embedding_model_name = embedding_model_name
 

qdrant_client/qdrant_fastembed.py

@@ -60,6 +60,7 @@ class QdrantFastembedMixin(QdrantBase):
     def __init__(self, **kwargs: Any):
         self._embedding_model_name: Optional[str] = None
         self._sparse_embedding_model_name: Optional[str] = None
+        self.local_files_only = kwargs.get("local_files_only", False)
         try:
             from fastembed import SparseTextEmbedding, TextEmbedding
 
@@ -119,11 +120,13 @@ def set_model(
                 stacklevel=2,
             )
 
+        local_files_only = kwargs.pop("local_files_only", self.local_files_only)
         self._get_or_init_model(
             model_name=embedding_model_name,
             cache_dir=cache_dir,
             threads=threads,
             providers=providers,
+            local_files_only=local_files_only,
             **kwargs,
         )
         self._embedding_model_name = embedding_model_name
@@ -134,6 +137,7 @@ def set_sparse_model(
         cache_dir: Optional[str] = None,
         threads: Optional[int] = None,
         providers: Optional[Sequence["OnnxProvider"]] = None,
+        **kwargs: Any,
     ) -> None:
         """
         Set sparse embedding model to use for hybrid search over documents in combination with dense embeddings.
@@ -154,12 +158,15 @@ def set_sparse_model(
         Returns:
             None
         """
+        local_files_only = kwargs.pop("local_files_only", self.local_files_only)
         if embedding_model_name is not None:
             self._get_or_init_sparse_model(
                 model_name=embedding_model_name,
                 cache_dir=cache_dir,
                 threads=threads,
                 providers=providers,
+                local_files_only=local_files_only,
+                **kwargs,
             )
         self._sparse_embedding_model_name = embedding_model_name
 

tests/test_fastembed.py

@@ -204,3 +204,23 @@ def test_idf_models():
 
     # the only sparse model without IDF is SPLADE, however it's too large for tests, so we don't test how non-idf
     # models work
+
+
+def test_local_files_only():
+    local_client = QdrantClient(":memory:")
+
+    if not local_client._FASTEMBED_INSTALLED:
+        pytest.skip("FastEmbed is not installed, skipping")
+
+    model_name = "sentence-transformers/all-MiniLM-L6-v2"
+    sparse_model_name = "Qdrant/bm42-all-minilm-l6-v2-attentions"
+
+    local_client.set_model(model_name)
+    local_client.set_sparse_model(sparse_model_name)
+
+    local_client.close()
+    local_client = QdrantClient(":memory:", local_files_only=True)
+    local_client.set_model(model_name)
+    local_client.set_sparse_model(sparse_model_name)
+    local_client.add("test_collection", **DOCS_EXAMPLE)
+    assert local_client.count("test_collection").count == 2