[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
update readme and 30 seconds example
yichuan520030910320c1bc2603a24e
diff --git a/README.md b/README.mdindex 003b469..08b9324 100755--- a/README.md+++ b/README.md@@ -68,27 +68,24 @@ uv sync```pythonfrom leann.api import LeannBuilder, LeannSearcher-# 1. Build index (no embeddings stored!)builder = LeannBuilder(backend_name="hnsw")+builder.add_text("C# is a powerful programming language")builder.add_text("Python is a powerful programming language")builder.add_text("Machine learning transforms industries")builder.add_text("Neural networks process complex data")builder.add_text("Leann is a great storage saving engine for RAG on your macbook")builder.build_index("knowledge.leann")-# 2. Search with real-time embeddingssearcher = LeannSearcher("knowledge.leann")-results = searcher.search("programming languages", top_k=2)--for result in results:- print(f"Score: {result['score']:.3f} - {result['text']}")+results = searcher.search("C++ programming languages", top_k=2, recompute_beighbor_embeddings=True)+print(results)```-### Run the Demo+### Run the Demo (support .pdf,.txt,.docx, .pptx, .csv, .md etc)```bash-uv run examples/document_search.py+uv run ./examples/main_cli_example.py```or you want to use python@@ -99,7 +96,7 @@ python ./examples/main_cli_example.py```**PDF RAG Demo (using LlamaIndex for document parsing and Leann for indexing/search)**-This demo showcases how to build a RAG system for PDF documents using Leann.+This demo showcases how to build a RAG system for PDF/md documents using Leann.1. Place your PDF files (and other supported formats like .docx, .pptx, .xlsx) into the `examples/data/` directory.2. Ensure you have an `OPENAI_API_KEY` set in your environment variables or in a `.env` file for the LLM to function.diff --git a/demo.ipynb b/demo.ipynbindex 0aacafe..c76f340 100644--- a/demo.ipynb+++ b/demo.ipynb@@ -2,361 +2,258 @@"cells": [{"cell_type": "code",- "execution_count": 1,+ "execution_count": 6,"metadata": {},"outputs": [+ {+ "name": "stderr",+ "output_type": "stream",+ "text": [+ "INFO:sentence_transformers.SentenceTransformer:Use pytorch device_name: mps\n",+ "INFO:sentence_transformers.SentenceTransformer:Load pretrained SentenceTransformer: facebook/contriever-msmarco\n",+ "WARNING:sentence_transformers.SentenceTransformer:No sentence-transformers model found with name facebook/contriever-msmarco. Creating a new one with mean pooling.\n"+ ]+ },{"name": "stdout","output_type": "stream","text": [- "Initializing leann-backend-diskann...\n",- "INFO: Registering backend 'diskann'\n",- "INFO: DiskANN backend loaded successfully\n",- "INFO: LeannBuilder initialized with 'diskann' backend.\n"+ "INFO: Computing embeddings for 1 chunks using SentenceTransformer model 'facebook/contriever-msmarco'...\n"]},{"name": "stderr","output_type": "stream","text": [- "/home/ubuntu/LEANN_clean/leann/.venv/lib/python3.11/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",- " from .autonotebook import tqdm as notebook_tqdm\n"+ "Batches: 100%|██████████| 1/1 [00:00<00:00, 65.35it/s]\n",+ "INFO:sentence_transformers.SentenceTransformer:Use pytorch device_name: mps\n",+ "INFO:sentence_transformers.SentenceTransformer:Load pretrained SentenceTransformer: facebook/contriever-msmarco\n",+ "WARNING:sentence_transformers.SentenceTransformer:No sentence-transformers model found with name facebook/contriever-msmarco. Creating a new one with mean pooling.\n"]},{"name": "stdout","output_type": "stream","text": [- "INFO: Computing embeddings for 6 chunks using 'sentence-transformers/all-mpnet-base-v2'...\n"+ "INFO: Computing embeddings for 5 chunks using SentenceTransformer model 'facebook/contriever-msmarco'...\n"]},{"name": "stderr","output_type": "stream","text": [- "Batches: 100%|██████████| 1/1 [00:00<00:00, 2.91it/s]\n"+ "Batches: 100%|██████████| 1/1 [00:00<00:00, 50.38it/s]"]},{"name": "stdout","output_type": "stream","text": [- "INFO: Building DiskANN index for 6 vectors with metric Metric.INNER_PRODUCT...\n",- "Using Inner Product search, so need to pre-process base data into temp file. Please ensure there is additional (n*(d+1)*4) bytes for storing pre-processed base vectors, apart from the interim indices created by DiskANN and the final index.\n",- "Pre-processing base file by adding extra coordinate\n",- "✅ DiskANN index built successfully at 'knowledge'\n",- "Writing bin: knowledge_disk.index_max_base_norm.bin\n",- "bin: #pts = 1, #dims = 1, size = 12B\n",- "Finished writing bin.\n",- "Time for preprocessing data for inner product: 0.000172 seconds\n",- "Reading max_norm_of_base from knowledge_disk.index_max_base_norm.bin\n",- "Reading bin file knowledge_disk.index_max_base_norm.bin ...\n",- "Opening bin file knowledge_disk.index_max_base_norm.bin... \n",- "Metadata: #pts = 1, #dims = 1...\n",- "done.\n",- "max_norm_of_base: 1\n",- "! Using prepped_base file at knowledge_prepped_base.bin\n",- "Starting index build: R=32 L=64 Query RAM budget: 4.02653e+09 Indexing ram budget: 8 T: 8\n",- "getting bin metadata\n",- "Time for getting bin metadata: 0.000019 seconds\n",- "Compressing 769-dimensional data into 512 bytes per vector.\n",- "Opened: knowledge_prepped_base.bin, size: 18464, cache_size: 18464\n",- "Training data with 6 samples loaded.\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 256, #dims = 769...\n",- "done.\n",- "PQ pivot file exists. Not generating again\n",- "Opened: knowledge_prepped_base.bin, size: 18464, cache_size: 18464\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 4, #dims = 1...\n",- "done.\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 256, #dims = 769...\n",- "done.\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 769, #dims = 1...\n",- "done.\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 513, #dims = 1...\n",- "done.\n",- "Loaded PQ pivot information\n",- "Processing points [0, 6)...done.\n",- "Time for generating quantized data: 0.055587 seconds\n",- "Full index fits in RAM budget, should consume at most 2.03973e-05GiBs, so building in one shot\n",- "L2: Using AVX2 distance computation DistanceL2Float\n",- "Passed, empty search_params while creating index config\n",- "Using only first 6 from file.. \n",- "Starting index build with 6 points... \n",- "0% of index build completed.Starting final cleanup..done. Link time: 0.00011s\n",- "Index built with degree: max:5 avg:5 min:5 count(deg<2):0\n",- "Not saving tags as they are not enabled.\n",- "Time taken for save: 0.000148s.\n",- "Time for building merged vamana index: 0.000836 seconds\n",- "Opened: knowledge_prepped_base.bin, size: 18464, cache_size: 18464\n",- "Vamana index file size=168\n",- "Opened: knowledge_disk.index, cache_size: 67108864\n",- "medoid: 0B\n",- "max_node_len: 3100B\n",- "nnodes_per_sector: 1B\n",- "# sectors: 6\n",- "Sector #0written\n",- "Finished writing 28672B\n",- "Writing bin: knowledge_disk.index\n",- "bin: #pts = 9, #dims = 1, size = 80B\n",- "Finished writing bin.\n",- "Output disk index file written to knowledge_disk.index\n",- "Finished writing 28672B\n",- "Time for generating disk layout: 0.040268 seconds\n",- "Opened: knowledge_prepped_base.bin, size: 18464, cache_size: 18464\n",- "Loading base knowledge_prepped_base.bin. #points: 6. #dim: 769.\n",- "Wrote 1 points to sample file: knowledge_sample_data.bin\n",- "Indexing time: 0.0970594\n",- "INFO: Leann metadata saved to knowledge.leann.meta.json\n"+ "M: 64 for level: 0\n",+ "INFO: Converting HNSW index to CSR-pruned format...\n",+ "Starting conversion: knowledge.index -> knowledge.csr.tmp\n",+ "[0.00s] Reading Index HNSW header...\n",+ "[0.00s] Header read: d=768, ntotal=5\n",+ "[0.00s] Reading HNSW struct vectors...\n",+ " Reading vector (dtype=<class 'numpy.float64'>, fmt='d')... Count=6, Bytes=48\n",+ "[0.00s] Read assign_probas (6)\n",+ " Reading vector (dtype=<class 'numpy.int32'>, fmt='i')... Count=7, Bytes=28\n",+ "[0.12s] Read cum_nneighbor_per_level (7)\n",+ " Reading vector (dtype=<class 'numpy.int32'>, fmt='i')... "]},{"name": "stderr","output_type": "stream","text": [- "Opened file : knowledge_disk.index\n"+ "\n"]},{"name": "stdout","output_type": "stream","text": [- "✅ DiskANN index loaded successfully.\n",- "INFO: LeannSearcher initialized with 'diskann' backend using index 'knowledge.leann'.\n",- "Since data is floating point, we assume that it has been appropriately pre-processed (normalization for cosine, and convert-to-l2 by adding extra dimension for MIPS). So we shall invoke an l2 distance function.\n",- "L2: Using AVX2 distance computation DistanceL2Float\n",- "L2: Using AVX2 distance computation DistanceL2Float\n",- "Before index load\n",- "Reading bin file knowledge_pq_compressed.bin ...\n",- "Opening bin file knowledge_pq_compressed.bin... \n",- "Metadata: #pts = 6, #dims = 512...\n",- "done.\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 4, #dims = 1...\n",- "done.\n",- "Offsets: 4096 791560 794644 796704\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 256, #dims = 769...\n",- "done.\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 769, #dims = 1...\n",- "done.\n",- "Reading bin file knowledge_pq_pivots.bin ...\n",- "Opening bin file knowledge_pq_pivots.bin... \n",- "Metadata: #pts = 513, #dims = 1...\n",- "done.\n",- "Loaded PQ Pivots: #ctrs: 256, #dims: 769, #chunks: 512\n",- "Loaded PQ centroids and in-memory compressed vectors. #points: 6 #dim: 769 #aligned_dim: 776 #chunks: 512\n",- "Loading index metadata from knowledge_disk.index\n",- "Disk-Index File Meta-data: # nodes per sector: 1, max node len (bytes): 3100, max node degree: 5\n",- "Disk-Index Meta: nodes per sector: 1, max node len: 3100, max node degree: 5\n",- "Setting up thread-specific contexts for nthreads: 8\n",- "allocating ctx: 0x7a33f7204000 to thread-id:134367072315200\n",- "allocating ctx: 0x7a33f6805000 to thread-id:134355206802368\n",- "allocating ctx: 0x7a33f5e72000 to thread-id:134355217288000\n",- "allocating ctx: 0x7a33f5e61000 to thread-id:134355227773632\n",- "allocating ctx: 0x7a33f5e50000 to thread-id:134355196316736\n",- "allocating ctx: 0x7a33f5e3f000 to thread-id:134355164859840\n",- "allocating ctx: 0x7a33f5e2e000 to thread-id:134355175345472\n",- "allocating ctx: 0x7a33f5e1d000 to thread-id:134355185831104\n",- "Loading centroid data from medoids vector data of 1 medoid(s)\n",- "Reading bin file knowledge_disk.index_max_base_norm.bin ...\n",- "Opening bin file knowledge_disk.index_max_base_norm.bin... \n",- "Metadata: #pts = 1, #dims = 1...\n",- "done.\n",- "Setting re-scaling factor of base vectors to 1\n",- "load_from_separate_paths done.\n",- "Reading (with alignment) bin file knowledge_sample_data.bin ...Metadata: #pts = 1, #dims = 769, aligned_dim = 776... allocating aligned memory of 3104 bytes... done. Copying data to mem_aligned buffer... done.\n",- "reserve ratio: 1\n",- "Graph traversal completed, hops: 3\n",- "Loading the cache list into memory....done.\n",- "After index load\n",- "INFO: Computing embeddings for 1 chunks using 'sentence-transformers/all-mpnet-base-v2'...\n"+ "Count=5, Bytes=20\n",+ "[0.21s] Read levels (5)\n",+ "[0.30s] Probing for compact storage flag...\n",+ "[0.30s] Found compact flag: False\n",+ "[0.30s] Compact flag is False, reading original format...\n",+ "[0.30s] Probing for potential extra byte before non-compact offsets...\n",+ "[0.30s] Found and consumed an unexpected 0x00 byte.\n",+ " Reading vector (dtype=<class 'numpy.uint64'>, fmt='Q')... Count=6, Bytes=48\n",+ "[0.30s] Read offsets (6)\n",+ "[0.39s] Attempting to read neighbors vector...\n",+ " Reading vector (dtype=<class 'numpy.int32'>, fmt='i')... Count=320, Bytes=1280\n",+ "[0.39s] Read neighbors (320)\n",+ "[0.47s] Read scalar params (ep=4, max_lvl=0)\n",+ "[0.47s] Checking for storage data...\n",+ "[0.47s] Found storage fourcc: 49467849.\n",+ "[0.47s] Converting to CSR format...\n",+ "[0.47s] Conversion loop finished. \n",+ "[0.47s] Running validation checks...\n",+ " Checking total valid neighbor count...\n",+ " OK: Total valid neighbors = 20\n",+ " Checking final pointer indices...\n",+ " OK: Final pointers match data size.\n",+ "[0.47s] Deleting original neighbors and offsets arrays...\n",+ " CSR Stats: |data|=20, |level_ptr|=10\n",+ "[0.56s] Writing CSR HNSW graph data in FAISS-compatible order...\n",+ " Pruning embeddings: Writing NULL storage marker.\n",+ "[0.64s] Conversion complete."]},{"name": "stderr","output_type": "stream","text": [- "Batches: 100%|██████████| 1/1 [00:00<00:00, 60.54it/s]"+ "INFO:sentence_transformers.SentenceTransformer:Use pytorch device_name: mps\n",+ "INFO:sentence_transformers.SentenceTransformer:Load pretrained SentenceTransformer: facebook/contriever-msmarco\n",+ "WARNING:sentence_transformers.SentenceTransformer:No sentence-transformers model found with name facebook/contriever-msmarco. Creating a new one with mean pooling.\n"]},{"name": "stdout","output_type": "stream","text": [- "INFO: DiskANN ZMQ mode enabled - ensuring embedding server is running\n",- "INFO: Starting session-level embedding server as a background process...\n",- "INFO: Running command from project root: /home/ubuntu/LEANN_clean/leann\n",- "INFO: Server process started with PID: 424761\n"+ "\n",+ "✅ CSR conversion successful.\n",+ "INFO: Replaced original index with CSR-pruned version at 'knowledge.index'\n",+ "[read_HNSW - CSR NL v4] Reading metadata & CSR indices (manual offset)...\n",+ "[read_HNSW NL v4] Read levels vector, size: 5\n",+ "[read_HNSW NL v4] Reading Compact Storage format indices...\n",+ "[read_HNSW NL v4] Read compact_level_ptr, size: 10\n",+ "INFO: Terminating session server process (PID: 21439)...\n",+ "[read_HNSW NL v4] Read compact_node_offsets, size: 6\n",+ "[read_HNSW NL v4] Read entry_point: 4, max_level: 0\n",+ "[read_HNSW NL v4] Read storage fourcc: 0x6c6c756e\n",+ "[read_HNSW NL v4 FIX] Detected FileIOReader. Neighbors size field offset: 326\n",+ "[read_HNSW NL v4] Reading neighbors data into memory.\n",+ "[read_HNSW NL v4] Read neighbors data, size: 20\n",+ "[read_HNSW NL v4] Finished reading metadata and CSR indices.\n",+ "INFO: Skipping external storage loading, since is_recompute is true.\n",+ "INFO: Server process terminated.\n",+ "🔍 DEBUG LeannSearcher.search() called:\n",+ " Query: 'C++ programming languages'\n",+ " Top_k: 2\n",+ " Search kwargs: {'recompute_beighbor_embeddings': True}\n",+ "INFO: Computing embeddings for 1 chunks using SentenceTransformer model 'facebook/contriever-msmarco'...\n"]},{"name": "stderr","output_type": "stream","text": [- "\n"+ "Batches: 100%|██████████| 1/1 [00:00<00:00, 85.08it/s]"]},{"name": "stdout","output_type": "stream","text": [- "✅ Embedding server is up and ready for this session.\n",- "[EmbeddingServer LOG]: Initializing leann-backend-diskann...\n",- "[EmbeddingServer LOG]: WARNING: Could not import DiskANN backend: cannot import name '_diskannpy' from partially initialized module 'packages.leann-backend-diskann.leann_backend_diskann' (most likely due to a circular import) (/home/ubuntu/LEANN_clean/leann/packages/leann-backend-diskann/leann_backend_diskann/__init__.py)\n",- "[EmbeddingServer LOG]: INFO: Initializing embedding server thread on port 5555\n",- "[EmbeddingServer LOG]: INFO: Using CUDA device\n",- "[EmbeddingServer LOG]: INFO: Loading model sentence-transformers/all-mpnet-base-v2\n",- "[EmbeddingServer LOG]: INFO: Using FP16 precision with model: sentence-transformers/all-mpnet-base-v2\n",- "[EmbeddingServer LOG]: INFO: Loaded 6 demo documents\n",- "[EmbeddingServer LOG]: INFO: ZMQ ROUTER server listening on port 5555\n",- "[EmbeddingServer LOG]: INFO: Embedding server ready to serve requests\n",- "[EmbeddingServer LOG]: INFO: Received ZMQ request from client 006b8b45, size 3 bytes\n",- "[EmbeddingServer LOG]: INFO: Request for 1 node embeddings: [0]\n",- "[EmbeddingServer LOG]: DEBUG: Node ID range: 0 to 0\n",- "[EmbeddingServer LOG]: Time taken for text lookup: 0.000028 seconds\n",- "[EmbeddingServer LOG]: INFO: Total batch size: 1, max_batch_size: 128\n",- "[EmbeddingServer LOG]: INFO: Processing batch of size 1\n",- "[EmbeddingServer LOG]: Time taken for tokenization (batch): 0.019294 seconds\n",- "[EmbeddingServer LOG]: Batch size: 1, Sequence length: 256\n",- "[EmbeddingServer LOG]: Time taken for transfer to device (batch): 0.000210 seconds\n",- "[EmbeddingServer LOG]: Time taken for embedding (batch): 3.065444 seconds\n",- "[EmbeddingServer LOG]: Time taken for mean pooling (batch): 0.041810 seconds\n",- "[EmbeddingServer LOG]: INFO: Serialize time: 0.000194 seconds\n",- "[EmbeddingServer LOG]: INFO: ZMQ E2E time: 3.128073 seconds\n",- "[EmbeddingServer LOG]: INFO: Received ZMQ request from client 006b8b45, size 7 bytes\n",- "[EmbeddingServer LOG]: INFO: Request for 5 node embeddings: [1, 2, 3, 4, 5]\n",- "[EmbeddingServer LOG]: DEBUG: Node ID range: 1 to 5\n",- "[EmbeddingServer LOG]: Time taken for text lookup: 0.000042 seconds\n",- "[EmbeddingServer LOG]: INFO: Total batch size: 5, max_batch_size: 128\n",- "[EmbeddingServer LOG]: INFO: Processing batch of size 5\n",- "[EmbeddingServer LOG]: Time taken for tokenization (batch): 0.001791 seconds\n",- "[EmbeddingServer LOG]: Batch size: 5, Sequence length: 256\n",- "[EmbeddingServer LOG]: Time taken for transfer to device (batch): 0.000112 seconds\n",- "[EmbeddingServer LOG]: Time taken for embedding (batch): 3.674183 seconds\n",- "[EmbeddingServer LOG]: Time taken for mean pooling (batch): 0.000372 seconds\n",- "[EmbeddingServer LOG]: INFO: Serialize time: 0.000177 seconds\n",- "[EmbeddingServer LOG]: INFO: ZMQ E2E time: 3.677425 seconds\n",- "[EmbeddingServer LOG]: INFO: Received ZMQ request from client 006b8b45, size 7 bytes\n",- "[EmbeddingServer LOG]: INFO: Request for 5 node embeddings: [3, 4, 2, 1, 0]\n",- "[EmbeddingServer LOG]: DEBUG: Node ID range: 0 to 4\n",- "[EmbeddingServer LOG]: Time taken for text lookup: 0.000030 seconds\n",- "[EmbeddingServer LOG]: INFO: Total batch size: 5, max_batch_size: 128\n",- "[EmbeddingServer LOG]: INFO: Processing batch of size 5\n",- "[EmbeddingServer LOG]: Time taken for tokenization (batch): 0.001550 seconds\n",- "[EmbeddingServer LOG]: Batch size: 5, Sequence length: 256\n",- "[EmbeddingServer LOG]: Time taken for transfer to device (batch): 0.000097 seconds\n",- "[EmbeddingServer LOG]: Time taken for embedding (batch): 0.009335 seconds\n",- "[EmbeddingServer LOG]: Time taken for mean pooling (batch): 0.000154 seconds\n",- "[EmbeddingServer LOG]: INFO: Serialize time: 0.000073 seconds\n",- "[EmbeddingServer LOG]: INFO: ZMQ E2E time: 0.011773 seconds\n",- "[EmbeddingServer LOG]: INFO: Received ZMQ request from client 006b8b45, size 7 bytes\n",- "[EmbeddingServer LOG]: INFO: Request for 5 node embeddings: [0, 1, 2, 4, 5]\n",- "[EmbeddingServer LOG]: DEBUG: Node ID range: 0 to 5\n",- "[EmbeddingServer LOG]: Time taken for text lookup: 0.000020 seconds\n",- "[EmbeddingServer LOG]: INFO: Total batch size: 5, max_batch_size: 128\n",- "[EmbeddingServer LOG]: INFO: Processing batch of size 5\n",- "[EmbeddingServer LOG]: Time taken for tokenization (batch): 0.001041 seconds\n",- "[EmbeddingServer LOG]: Batch size: 5, Sequence length: 256\n",- "[EmbeddingServer LOG]: Time taken for transfer to device (batch): 0.000125 seconds\n",- "[EmbeddingServer LOG]: Time taken for embedding (batch): 0.008972 seconds\n",- "[EmbeddingServer LOG]: Time taken for mean pooling (batch): 0.000151 seconds\n",- "[EmbeddingServer LOG]: INFO: Serialize time: 0.000048 seconds\n",- "[EmbeddingServer LOG]: INFO: ZMQ E2E time: 0.010853 seconds\n",- "[EmbeddingServer LOG]: INFO: Received ZMQ request from client 006b8b45, size 7 bytes\n",- "[EmbeddingServer LOG]: INFO: Request for 5 node embeddings: [3, 1, 0, 2, 5]\n",- "[EmbeddingServer LOG]: DEBUG: Node ID range: 0 to 5\n",- "[EmbeddingServer LOG]: Time taken for text lookup: 0.000020 seconds\n",- "[EmbeddingServer LOG]: INFO: Total batch size: 5, max_batch_size: 128\n",- "[EmbeddingServer LOG]: INFO: Processing batch of size 5\n",- "[EmbeddingServer LOG]: Time taken for tokenization (batch): 0.001350 seconds\n",- "[EmbeddingServer LOG]: Batch size: 5, Sequence length: 256\n",- "[EmbeddingServer LOG]: Time taken for transfer to device (batch): 0.000088 seconds\n",- "[EmbeddingServer LOG]: Time taken for embedding (batch): 0.008869 seconds\n",- "[EmbeddingServer LOG]: Time taken for mean pooling (batch): 0.000146 seconds\n",- "[EmbeddingServer LOG]: INFO: Serialize time: 0.000063 seconds\n",- "[EmbeddingServer LOG]: INFO: ZMQ E2E time: 0.011054 seconds\n",- "[EmbeddingServer LOG]: INFO: Received ZMQ request from client 006b8b45, size 7 bytes\n",- "[EmbeddingServer LOG]: INFO: Request for 5 node embeddings: [0, 2, 3, 4, 5]\n",- "[EmbeddingServer LOG]: DEBUG: Node ID range: 0 to 5\n",- "[EmbeddingServer LOG]: Time taken for text lookup: 0.000022 seconds\n",- "[EmbeddingServer LOG]: INFO: Total batch size: 5, max_batch_size: 128\n",- "[EmbeddingServer LOG]: INFO: Processing batch of size 5\n",- "[EmbeddingServer LOG]: Time taken for tokenization (batch): 0.001195 seconds\n",- "[EmbeddingServer LOG]: Batch size: 5, Sequence length: 256\n",- "[EmbeddingServer LOG]: Time taken for transfer to device (batch): 0.000087 seconds\n",- "[EmbeddingServer LOG]: Time taken for embedding (batch): 0.008903 seconds\n",- "[EmbeddingServer LOG]: Time taken for mean pooling (batch): 0.000145 seconds\n",- "[EmbeddingServer LOG]: INFO: Serialize time: 0.000060 seconds\n",- "[EmbeddingServer LOG]: INFO: ZMQ E2E time: 0.010921 seconds\n",- "[EmbeddingServer LOG]: INFO: Received ZMQ request from client 006b8b45, size 7 bytes\n",- "[EmbeddingServer LOG]: INFO: Request for 5 node embeddings: [1, 0, 3, 4, 5]\n",- "[EmbeddingServer LOG]: DEBUG: Node ID range: 0 to 5\n",- "[EmbeddingServer LOG]: Time taken for text lookup: 0.000020 seconds\n",- "[EmbeddingServer LOG]: INFO: Total batch size: 5, max_batch_size: 128\n",- "[EmbeddingServer LOG]: INFO: Processing batch of size 5\n",- "[EmbeddingServer LOG]: Time taken for tokenization (batch): 0.001188 seconds\n",- "[EmbeddingServer LOG]: Batch size: 5, Sequence length: 256\n",- "[EmbeddingServer LOG]: Time taken for transfer to device (batch): 0.000087 seconds\n",- "[EmbeddingServer LOG]: Time taken for embedding (batch): 0.008858 seconds\n",- "[EmbeddingServer LOG]: Time taken for mean pooling (batch): 0.000153 seconds\n",- "[EmbeddingServer LOG]: INFO: Serialize time: 0.000052 seconds\n",- "[EmbeddingServer LOG]: INFO: ZMQ E2E time: 0.010886 seconds\n",- "reserve ratio: Score: -0.481 - C++ is a powerful programming language1\n",- "Graph traversal completed, hops: 3\n",+ " Generated embedding shape: (1, 768)\n",+ "🔍 DEBUG Query embedding first 10 values: [ 0.04288 -0.04135 0.0666 0.02197 -0.0881 -0.04367 -0.02835 -0.0408\n",+ " -0.1254 -0.08594]\n",+ "🔍 DEBUG Query embedding norm: 1.3876953125\n",+ "INFO: Starting session-level embedding server for 'leann_backend_hnsw.hnsw_embedding_server'...\n",+ "INFO: Running command from project root: /Users/yichuan/Desktop/code/LEANN/leann\n",+ "INFO: Command: /Users/yichuan/Desktop/code/LEANN/leann/.venv/bin/python -m leann_backend_hnsw.hnsw_embedding_server --zmq-port 5557 --model-name facebook/contriever-msmarco --passages-file knowledge.leann.meta.json --disable-warmup\n",+ "INFO: Server process started with PID: 21622\n"+ ]+ },+ {+ "name": "stderr",+ "output_type": "stream",+ "text": ["\n",- "Score: -1.049 - Java is a powerful programming language\n"+ "huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...\n",+ "To disable this warning, you can either:\n",+ "\t- Avoid using `tokenizers` before the fork if possible\n",+ "\t- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)\n"+ ]+ },+ {+ "name": "stdout",+ "output_type": "stream",+ "text": [+ "✅ Embedding server is up and ready for this session.\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: INFO: Starting backend auto-discovery...\n",+ "ZmqDistanceComputer initialized: d=768, metric=0\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: INFO: Registering backend 'diskann'\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: INFO: Backend auto-discovery finished.\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: INFO: Registering backend 'hnsw'\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Loading tokenizer for facebook/contriever-msmarco...\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Tokenizer loaded successfully!\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: MPS available: True\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: CUDA available: False\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Using MPS device (Apple Silicon)\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Starting HNSW server on port 5557 with model facebook/contriever-msmarco\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Loading model facebook/contriever-msmarco... (this may take a while if downloading)\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Model facebook/contriever-msmarco loaded successfully!\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Loaded label map with 5 entries\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Initialized lazy passage loading for 5 passages\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Warmup disabled or no passages available (enable_warmup=False, passages=5)\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: HNSW ZMQ server listening on port 5557\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Started HNSW ZMQ server thread on port 5557\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Received ZMQ request of size 3 bytes\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: request_payload length: 1\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: request_payload[0]: <class 'list'> - [4]\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Request for 1 node embeddings\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Total batch size: 1, max_batch_size: 128\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG zmq_server_thread: Final 'hidden' array | Shape: (1, 768) | Dtype: float32 | Has NaN/Inf: False\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Serialize time: 0.000150 seconds\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: ZMQ E2E time: 0.142946 seconds\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Received ZMQ request of size 3849 bytes\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: request_payload length: 2\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: request_payload[0]: <class 'list'> - [0, 1, 2, 3]\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: request_payload[1]: <class 'list'> - [0.042877197265625, -0.041351318359375, 0.06658935546875, 0.02197265625, -0.08807373046875, -0.04367...\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Distance calculation request received\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Node IDs: [0, 1, 2, 3]\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Query vector dim: 768\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Passages loaded: 5\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Looking up passage ID 0\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Found text for ID 0, length: 37\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Looking up passage ID 1\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Found text for ID 1, length: 41\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Looking up passage ID 2\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Found text for ID 2, length: 38\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Looking up passage ID 3\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: DEBUG: Found text for ID 3, length: 36\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Sending distance response with 4 distances\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: Distance calculation E2E time: 0.173929 seconds\n",+ " Backend returned: labels=2 results\n",+ " Processing 2 passage IDs:\n",+ " 1. passage_id='28b7b6b9-d0a4-408d-9e7f-9a7fcb7d8186' -> SUCCESS: C# is a powerful programming language...\n",+ " 2. passage_id='1bccf691-a571-4e9d-aaed-424a30ba8604' -> SUCCESS: Python is a powerful programming language...\n",+ " Final enriched results: 2 passages\n",+ "[SearchResult(id='28b7b6b9-d0a4-408d-9e7f-9a7fcb7d8186', score=np.float32(1.5213046), text='C# is a powerful programming language', metadata={}), SearchResult(id='1bccf691-a571-4e9d-aaed-424a30ba8604', score=np.float32(1.2999034), text='Python is a powerful programming language', metadata={})]\n"]},{"name": "stdout","output_type": "stream","text": [- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n",- "[EmbeddingServer LOG]: INFO: ZMQ socket timeout, continuing to listen\n"+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: ZMQ socket timeout, continuing to listen\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: ZMQ socket timeout, continuing to listen\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: ZMQ socket timeout, continuing to listen\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: ZMQ socket timeout, continuing to listen\n",+ "[leann_backend_hnsw.hnsw_embedding_server LOG]: ZMQ socket timeout, continuing to listen\n"]}],"source": ["from leann.api import LeannBuilder, LeannSearcher\n",- "import leann_backend_diskann\n","# 1. Build index (no embeddings stored!)\n",- "builder = LeannBuilder(backend_name=\"diskann\")\n",+ "builder = LeannBuilder(backend_name=\"hnsw\")\n",+ "builder.add_text(\"C# is a powerful programming language\")\n","builder.add_text(\"Python is a powerful programming language\")\n","builder.add_text(\"Machine learning transforms industries\") \n","builder.add_text(\"Neural networks process complex data\")\n",- "builder.add_text(\"Java is a powerful programming language\")\n",- "builder.add_text(\"C++ is a powerful programming language\")\n",- "builder.add_text(\"C# is a powerful programming language\")\n",+ "builder.add_text(\"Leann is a great storage saving engine for RAG on your macbook\")\n","builder.build_index(\"knowledge.leann\")\n",- "\n","# 2. Search with real-time embeddings\n","searcher = LeannSearcher(\"knowledge.leann\")\n",- "results = searcher.search(\"C++ programming languages\", top_k=2,recompute_beighbor_embeddings=True)\n",- "\n",- "for result in results:\n",- " print(f\"Score: {result['score']:.3f} - {result['text']}\")"+ "results = searcher.search(\"C++ programming languages\", top_k=2, recompute_beighbor_embeddings=True)\n",+ "print(results)"]}],@@ -376,7 +273,7 @@"name": "python","nbconvert_exporter": "python","pygments_lexer": "ipython3",- "version": "3.11.11"+ "version": "3.11.12"}},"nbformat": 4,