omendb 0.0.25

Fast embedded vector database with HNSW + ACORN-1 filtered search

OmenDB

Embedded vector database for Python and Node.js. No server, no setup, just install.

pip install omendb

Quick Start

import omendb

# Create database (persistent) - creates ./mydb.omen file
db = omendb.open("./mydb", dimensions=128)

# Add vectors with metadata
db.set([
    {"id": "doc1", "vector": [0.1] * 128, "metadata": {"category": "science"}},
    {"id": "doc2", "vector": [0.2] * 128, "metadata": {"category": "history"}},
])

# Search
results = db.search([0.1] * 128, k=5)

# Filtered search
results = db.search([0.1] * 128, k=5, filter={"category": "science"})

Features

• Embedded - Runs in-process, no server needed
• Persistent - Data survives restarts automatically
• Filtered search - Query by metadata with JSON-style filters
• Hybrid search - Combine vector similarity with BM25 text search
• Quantization - 4x smaller indexes with minimal recall loss

Platforms

Platform	Status
Linux (x86_64, ARM64)	Supported
macOS (Intel, Apple Silicon)	Supported
Windows (x86_64)	Experimental

API

# Database
db = omendb.open(path, dimensions)      # Open or create
db = omendb.open(":memory:", dimensions)  # In-memory (ephemeral)

# CRUD
db.set(items)                           # Insert/update vectors
db.get(id)                              # Get by ID
db.get_batch(ids)                       # Batch get by IDs
db.delete(ids)                          # Delete by IDs
db.delete_by_filter(filter)             # Delete by metadata filter
db.update(id, metadata)                 # Update metadata only

# Iteration
len(db)                                 # Number of vectors
db.count()                              # Same as len(db)
db.count(filter={...})                  # Count matching filter
db.ids()                                # Iterate all IDs (lazy)
db.items()                              # Get all items as list
db.exists(id)                           # Check if ID exists
"id" in db                              # Same as exists()
for item in db: ...                     # Iterate all items (lazy)

# Search
db.search(query, k)                     # Vector search
db.search(query, k, filter={...})       # Filtered search
db.search(query, k, max_distance=0.5)   # Only results with distance <= 0.5
db.search_batch(queries, k)             # Batch search (parallel)

# Hybrid search (requires text field in vectors)
db.search_hybrid(query_vector, query_text, k)
db.search_hybrid(query_vector, query_text, k, alpha=0.7)  # 70% vector, 30% text
db.search_hybrid(query_vector, query_text, k, subscores=True)  # Return separate scores
db.search_text(query_text, k)           # Text-only BM25

# Persistence
db.flush()                              # Flush to disk
db.close()                              # Close and release file locks
db.compact()                            # Remove deleted records, reclaim space
db.optimize()                           # Reorder graph for cache locality

Distance Filtering

Use max_distance to filter out low-relevance results (prevents "context rot" in RAG):

# Only return results with distance <= 0.5
results = db.search(query, k=10, max_distance=0.5)

# Combine with metadata filter
results = db.search(query, k=10, filter={"type": "doc"}, max_distance=0.5)

This ensures your RAG pipeline only receives highly relevant context, avoiding distractors that can hurt LLM performance.

Filters

# Equality
{"field": "value"}                      # Shorthand
{"field": {"$eq": "value"}}             # Explicit

# Comparison
{"field": {"$ne": "value"}}             # Not equal
{"field": {"$gt": 10}}                  # Greater than
{"field": {"$gte": 10}}                 # Greater or equal
{"field": {"$lt": 10}}                  # Less than
{"field": {"$lte": 10}}                 # Less or equal

# Membership
{"field": {"$in": ["a", "b"]}}          # In list
{"field": {"$contains": "sub"}}         # String contains

# Logical
{"$and": [{...}, {...}]}                # AND
{"$or": [{...}, {...}]}                 # OR

Configuration

db = omendb.open(
    "./mydb",              # Creates ./mydb.omen + ./mydb.wal
    dimensions=384,
    m=16,                # HNSW connections per node (default: 16)
    ef_construction=200, # Index build quality (default: 100)
    ef_search=100,       # Search quality (default: 100)
    quantization=True,   # SQ8 quantization (default: None)
    metric="cosine",     # Distance metric (default: "l2")
)

# Quantization options:
# - True or "sq8": SQ8 ~4x smaller, ~99% recall (recommended)
# - None/False: Full precision (default)

# Distance metric options:
# - "l2" or "euclidean": Euclidean distance (default)
# - "cosine": Cosine distance (1 - cosine similarity)
# - "dot" or "ip": Inner product (for MIPS)

# Context manager (auto-flush on exit)
with omendb.open("./db", dimensions=768) as db:
    db.set([...])

# Hybrid search with alpha (0=text, 1=vector, default=0.5)
db.search_hybrid(query_vec, "query text", k=10, alpha=0.7)

# Get separate keyword and semantic scores for debugging/tuning
results = db.search_hybrid(query_vec, "query text", k=10, subscores=True)
# Returns: {"id": "...", "score": 0.85, "keyword_score": 0.92, "semantic_score": 0.78}

Performance

10K vectors, Apple M3 Max (m=16, ef=100, k=10). Measured 2026-01-20:

Dimension	Single QPS	Batch QPS	Speedup
128D	11,542	82,015	7.1x
768D	3,531	26,254	7.4x
1536D	1,825	7,579	4.2x

SIFT-1M (1M vectors, 128D, m=16, ef=100, k=10):

Machine	QPS	Recall
i9-13900KF	4,591	98.6%
Apple M3 Max	3,216	98.4%

Quantization reduces memory with minimal recall loss:

Mode	Compression	Use Case
f32	1x	Default, highest recall
sq8	4x	Recommended for most users

db = omendb.open("./db", dimensions=768, quantization=True)  # Enable SQ8

Benchmark methodology

• Parameters: m=16, ef_construction=100, ef_search=100
• Batch: Uses Rayon for parallel search across all cores
• Recall: Validated against brute-force ground truth on SIFT/GloVe
• Reproduce:
  • Quick (10K): uv run python benchmarks/run.py
  • SIFT-1M: uv run python benchmarks/ann_dataset_test.py --dataset sift-128-euclidean

Tuning ef_search for High Dimensions

The ef_search parameter controls the recall/speed tradeoff at query time. Higher values explore more candidates, improving recall but slowing search.

Rules of thumb:

• ef_search must be >= k (number of results requested)
• For 128D embeddings: ef=100 usually achieves 90%+ recall
• For 768D+ embeddings: increase to ef=200-400 for better recall
• If recall drops at scale (50K+), increase both ef_search and ef_construction

Runtime tuning:

# Check current value
print(db.get_ef_search())  # 100

# Increase for better recall (slower)
db.set_ef_search(200)

# Decrease for speed (may reduce recall)
db.set_ef_search(50)

# Per-query override
results = db.search(query, k=10, ef=300)

Recommended settings by use case:

Use Case	ef_search	Expected Recall
Fast search (128D)	64	~85%
Balanced (default)	100	~90%
High recall (768D+)	200-300	~95%+
Maximum recall	500+	~98%+

Examples

See complete working examples:

• python/examples/quickstart.py - Minimal Python example
• python/examples/basic.py - CRUD operations and persistence
• python/examples/filters.py - All filter operators
• python/examples/rag.py - RAG workflow with mock embeddings
• node/examples/quickstart.js - Minimal Node.js example

Integrations

LangChain

pip install omendb[langchain]

from langchain_openai import OpenAIEmbeddings
from omendb.langchain import OmenDBVectorStore

store = OmenDBVectorStore.from_texts(
    texts=["Paris is the capital of France"],
    embedding=OpenAIEmbeddings(),
    path="./langchain_vectors",
)
docs = store.similarity_search("capital of France", k=1)

LlamaIndex

pip install omendb[llamaindex]

from llama_index.core import VectorStoreIndex, Document, StorageContext
from omendb.llamaindex import OmenDBVectorStore

vector_store = OmenDBVectorStore(path="./llama_vectors")
storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(
    [Document(text="OmenDB is fast")],
    storage_context=storage_context,
)
response = index.as_query_engine().query("What is OmenDB?")

License

Elastic License 2.0 - Free to use, modify, and embed. The only restriction: you can't offer OmenDB as a managed service to third parties.