[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: more logs
Andy Leeab72a2ab9db7
diff --git a/packages/leann-core/src/leann/api.py b/packages/leann-core/src/leann/api.pyindex 879708c..cbee3f7 100644--- a/packages/leann-core/src/leann/api.py+++ b/packages/leann-core/src/leann/api.py@@ -14,6 +14,9 @@ from dataclasses import dataclass, fieldfrom .registry import BACKEND_REGISTRYfrom .interface import LeannBackendFactoryInterfacefrom .chat import get_llm+import logging++logger = logging.getLogger(__name__)def compute_embeddings(@@ -67,8 +70,8 @@ def compute_embeddings_via_server(chunks: List of text chunks to embedmodel_name: Name of the sentence transformer model"""- print(- f"INFO: Computing embeddings for {len(chunks)} chunks using SentenceTransformer model '{model_name}' (via embedding server)..."+ logger.info(+ f"Computing embeddings for {len(chunks)} chunks using SentenceTransformer model '{model_name}' (via embedding server)...")import zmqimport msgpack@@ -288,7 +291,7 @@ class LeannBuilder:f"Dimension mismatch: expected {self.dimensions}, got {embedding_dim}")- print(+ logger.info(f"Building index from precomputed embeddings: {len(ids)} items, {embedding_dim} dimensions")@@ -296,7 +299,7 @@ class LeannBuilder:if len(self.chunks) != len(ids):# If no text chunks provided, create placeholder text entriesif not self.chunks:- print("No text chunks provided, creating placeholder entries...")+ logger.info("No text chunks provided, creating placeholder entries...")for id_val in ids:self.add_text(f"Document {id_val}",@@ -371,7 +374,9 @@ class LeannBuilder:with open(leann_meta_path, "w", encoding="utf-8") as f:json.dump(meta_data, f, indent=2)- print(f"Index built successfully from precomputed embeddings: {index_path}")+ logger.info(+ f"Index built successfully from precomputed embeddings: {index_path}"+ )class LeannSearcher:@@ -411,10 +416,10 @@ class LeannSearcher:expected_zmq_port: int = 5557,**kwargs,) -> List[SearchResult]:- print("🔍 DEBUG LeannSearcher.search() called:")- print(f" Query: '{query}'")- print(f" Top_k: {top_k}")- print(f" Additional kwargs: {kwargs}")+ logger.info("🔍 LeannSearcher.search() called:")+ logger.info(f" Query: '{query}'")+ logger.info(f" Top_k: {top_k}")+ logger.info(f" Additional kwargs: {kwargs}")start_time = time.time()@@ -432,9 +437,9 @@ class LeannSearcher:use_server_if_available=recompute_embeddings,zmq_port=zmq_port,)- print(f" Generated embedding shape: {query_embedding.shape}")+ logger.info(f" Generated embedding shape: {query_embedding.shape}")embedding_time = time.time() - start_time- print(f" Embedding time: {embedding_time} seconds")+ logger.info(f" Embedding time: {embedding_time} seconds")start_time = time.time()results = self.backend_impl.search(@@ -449,14 +454,14 @@ class LeannSearcher:**kwargs,)search_time = time.time() - start_time- print(f" Search time: {search_time} seconds")- print(+ logger.info(f" Search time: {search_time} seconds")+ logger.info(f" Backend returned: labels={len(results.get('labels', [[]])[0])} results")enriched_results = []if "labels" in results and "distances" in results:- print(f" Processing {len(results['labels'][0])} passage IDs:")+ logger.info(f" Processing {len(results['labels'][0])} passage IDs:")for i, (string_id, dist) in enumerate(zip(results["labels"][0], results["distances"][0])):@@ -470,15 +475,15 @@ class LeannSearcher:metadata=passage_data.get("metadata", {}),))- print(+ logger.info(f" {i + 1}. passage_id='{string_id}' -> SUCCESS: {passage_data['text']}...")except KeyError:- print(+ logger.error(f" {i + 1}. passage_id='{string_id}' -> ERROR: Passage not found in PassageManager!")- print(f" Final enriched results: {len(enriched_results)} passages")+ logger.info(f" Final enriched results: {len(enriched_results)} passages")return enriched_resultsdiff --git a/packages/leann-core/src/leann/embedding_compute.py b/packages/leann-core/src/leann/embedding_compute.pyindex d0df463..ee2a59d 100644--- a/packages/leann-core/src/leann/embedding_compute.py+++ b/packages/leann-core/src/leann/embedding_compute.py@@ -8,8 +8,13 @@ import numpy as npimport torchfrom typing import List, Dict, Anyimport logging+import os+# Set up logger with proper levellogger = logging.getLogger(__name__)+LOG_LEVEL = os.getenv("LEANN_LOG_LEVEL", "WARNING").upper()+log_level = getattr(logging, LOG_LEVEL, logging.WARNING)+logger.setLevel(log_level)# Global model cache to avoid repeated loading_model_cache: Dict[str, Any] = {}@@ -125,7 +130,9 @@ def compute_embeddings_sentence_transformers(try:model = model.half()model = torch.compile(model)- logger.info(f"Using FP16 precision and compile optimization: {model_name}")+ logger.info(+ f"Using FP16 precision and compile optimization: {model_name}"+ )except Exception as e:logger.warning(f"FP16 or compile optimization failed: {e}")@@ -145,8 +152,8 @@ def compute_embeddings_sentence_transformers(device=device,)- print(- f"INFO: Generated {len(embeddings)} embeddings, dimension: {embeddings.shape[1]}"+ logger.info(+ f"Generated {len(embeddings)} embeddings, dimension: {embeddings.shape[1]}")# Validate results@@ -212,8 +219,8 @@ def compute_embeddings_openai(texts: List[str], model_name: str) -> np.ndarray:raiseembeddings = np.array(all_embeddings, dtype=np.float32)- print(- f"INFO: Generated {len(embeddings)} embeddings, dimension: {embeddings.shape[1]}"+ logger.info(+ f"Generated {len(embeddings)} embeddings, dimension: {embeddings.shape[1]}")return embeddings