[MLsys2026 Best Paper]: https://arxiv.org/abs/2506.08276. RAG on Everything with LEANN. Enjoy 97% storage savings while running a fast, accurate, and 100% private RAG application on your personal device.
Python aifaissgpt-osslangchainllama-indexllmlocalstorageoffline-firstollamaprivacypythonragretrieval-augmented-generationvector-databasevector-searchvectors
fix: mlx
Andy Leec112956d2dab
diff --git a/packages/leann-backend-hnsw/leann_backend_hnsw/hnsw_embedding_server.py b/packages/leann-backend-hnsw/leann_backend_hnsw/hnsw_embedding_server.pyindex 3d3c9fa..b20959b 100644--- a/packages/leann-backend-hnsw/leann_backend_hnsw/hnsw_embedding_server.py+++ b/packages/leann-backend-hnsw/leann_backend_hnsw/hnsw_embedding_server.py@@ -57,7 +57,6 @@ def create_hnsw_embedding_server(finally:sys.path.pop(0)-# Check port availabilityimport socket@@ -152,7 +151,9 @@ def create_hnsw_embedding_server(raise# Process embeddings- embeddings = compute_embeddings(texts, model_name, mode=embedding_mode)+ embeddings = compute_embeddings(+ texts, model_name, mode=embedding_mode+ )print(f"INFO: Computed embeddings for {len(texts)} texts, shape: {embeddings.shape}")@@ -280,7 +281,7 @@ if __name__ == "__main__":"--embedding-mode",type=str,default="sentence-transformers",- choices=["sentence-transformers", "openai"],+ choices=["sentence-transformers", "openai", "mlx"],help="Embedding backend mode",)diff --git a/packages/leann-core/src/leann/searcher_base.py b/packages/leann-core/src/leann/searcher_base.pyindex 73f979a..8592ed4 100644--- a/packages/leann-core/src/leann/searcher_base.py+++ b/packages/leann-core/src/leann/searcher_base.py@@ -42,8 +42,10 @@ class BaseSearcher(LeannBackendSearcherInterface, ABC):"WARNING: embedding_model not found in meta.json. Recompute will fail.")+ self.embedding_mode = self.meta.get("embedding_mode", "sentence-transformers")+self.embedding_server_manager = EmbeddingServerManager(- backend_module_name=backend_module_name+ backend_module_name=backend_module_name,)def _load_meta(self) -> Dict[str, Any]:@@ -67,14 +69,12 @@ class BaseSearcher(LeannBackendSearcherInterface, ABC):"Cannot use recompute mode without 'embedding_model' in meta.json.")- embedding_mode = self.meta.get("embedding_mode", "sentence-transformers")-server_started, actual_port = self.embedding_server_manager.start_server(port=port,model_name=self.embedding_model,+ embedding_mode=self.embedding_mode,passages_file=passages_source_file,distance_metric=kwargs.get("distance_metric"),- embedding_mode=embedding_mode,enable_warmup=kwargs.get("enable_warmup", False),)if not server_started:diff --git a/test/build_mlx_index.py b/test/build_mlx_index.pyindex 502cc24..a76bdee 100644--- a/test/build_mlx_index.py+++ b/test/build_mlx_index.py@@ -12,7 +12,7 @@ else:builder = LeannBuilder(backend_name="hnsw",embedding_model="mlx-community/Qwen3-Embedding-0.6B-4bit-DWQ",- use_mlx=True,+ embedding_mode="mlx",)# 2. Add documents