ShibuDb

ShibuDb is a lightweight database system with vector search capabilities powered by FAISS. It provides high-performance storage and retrieval with support for both traditional key-value operations and advanced vector similarity search.

• 🔍 Vector Search: Advanced similarity search using FAISS
• 🏷️ Metadata Filtering: Pre-filter vector search by indexed metadata fields on Flat spaces (--where filters)
• 🗄️ Multi-Space Architecture: Organize data into separate spaces
• 🔐 Role-Based Access Control: Secure authentication and authorization
• ⚡ High Performance: Optimized storage with B-tree indexing
• 🌐 Cross-Platform: Linux (AMD64/ARM64) and macOS (AMD64/ARM64)
• 📊 Dynamic Connection Management: Runtime connection limit updates
• 🛡️ Data Durability: Write-Ahead Logging for crash recovery

Benchmarked on v1.2.0 with SIFT1M (single 16-vCPU node): ~62k ops/s key-value reads/writes at 1M keys, and sub-millisecond HNSW vector search (0.85 recall@10, ~4,600 QPS peak at 1M vectors). See the full Performance Benchmarks for all index types, scales, and WAL cost analysis.

# From source
git clone https://github.com/shibudb-org/shibudb-server.git
cd shibudb-server

# Start the local server on port 4444.
# This dev script bootstraps an admin user as admin:admin (see scripts/start-local-server.sh).
make start-local-server

# Connect to database on default 4444 port
make connect-local-client

# Create your first space with engine type key-value
CREATE-SPACE my_data --engine key-value

# Switch to the created space
USE my_data

# Store and retrieve data
PUT user:1 "John Doe"
GET user:1

• Setup Guide - Complete installation and configuration guide
• Architecture - System architecture and design

• Key-Value Engine - Comprehensive guide to key-value operations
• Vector Engine - Vector search capabilities and FAISS integration
• User Management - Authentication, roles, and permissions

• Dynamic Connection Limiting - Runtime connection management
• Administration Guide - Server administration and monitoring

• API Reference - Complete command reference
• Performance Benchmarks - Measured throughput, latency, and recall across all engines and index types
• Troubleshooting - Common issues and solutions

ShibuDb follows a modular architecture with clear separation of concerns:

• Storage Engine: Efficient key-value and vector storage with WAL
• Query Engine: Processes and executes database operations
• Authentication: Role-based access control system
• Space Management: Multi-tenant data organization
• Indexing: B-tree and FAISS vector indexes for fast retrieval
• Connection Management: Dynamic connection limiting and monitoring
• Management API: HTTP endpoints for runtime control

• Go 1.24.0 or later
• FAISS libraries (included in resources/)

# Run tests
make test

# Run benchmarks
make benchmark

# Run E2E tests, this requires test server (make start-local-server)
make e2e-test

For local development and testing, you can use the following commands:

# Start the local development server (port 4444)
make start-local-server

# Connect to the local server using the CLI client
make connect-local-client

Default credentials for local development (when using make start-local-server):

• Username: admin
• Password: admin
• Port: 4444

Available CLI Commands:

• help or ? - Show the full interactive command reference
• USE <space> - Switch to a specific space
• create-space <name> [--engine key-value|vector] [--dimension N] [--index-type TYPE] [--metric METRIC] [--metadata-fields name:type,...] [--compression TurboQuant2Bits|TurboQuant3Bits|TurboQuant4Bits] - Create a new space
• put <key> <value> - Store a key-value pair
• get <key> - Retrieve a value by key
• delete <key> - Delete a key-value pair
• insert-vector <id> <comma-separated-floats> [--meta key=value,...] - Insert a vector (optionally with indexed metadata)
• delete-vector <id> - Delete a vector (Not supported for index type HNSW)
• search-topk <comma-separated-floats> <k> [--where <expression>] - Search for top-k similar vectors (optionally pre-filtered by metadata)
• range-search <comma-separated-floats> <radius> [--where <expression>] - Search within a radius (optionally pre-filtered by metadata)
• create-user - Create a new user (admin only)
• list-spaces - List all available spaces
• exit or quit - Exit the client

Example Workflow:

# Terminal 1: Start the server
make start-local-server

# Terminal 2: Connect and interact
make connect-local-client

# In the client:
[admin]> create-space mydata
[admin]> USE mydata
[mydata]> put key1 value1
[mydata]> get key1
[mydata]> create-space vectors --engine vector --dimension 128
[mydata]> USE vectors
[vectors]> insert-vector vec1 1.0,2.0,3.0,4.0
[vectors]> search-topk 1.1,2.1,3.1,4.1 5

Metadata Filtering (Flat index):

Declare indexed metadata fields when creating a Flat vector space, attach metadata on insert, then pre-filter searches with a --where expression. This restricts the similarity search to matching vectors (e.g. per-tenant data) for better recall and speed.

# Declare indexed fields (name:type, types: string | int | float)
[admin]> create-space products --engine vector --dimension 4 --index-type Flat --metric L2 \
            --metadata-fields user_id:string,category:string,price:float,year:int
[admin]> USE products

# Attach metadata on insert (key=value; numeric values are inferred, quote to force a string)
[products]> insert-vector 1 0.1,0.1,0.1,0.1 --meta user_id=alice,category=books,price=12.5,year=2020
[products]> insert-vector 2 0.2,0.2,0.2,0.2 --meta user_id=bob,category=books,price=40,year=2022
[products]> insert-vector 3 0.15,0.15,0.15,0.15 --meta user_id=alice,category=toys,price=5,year=2023

# Filtered search
[products]> search-topk 0.1,0.1,0.1,0.1 10 --where user_id=alice
[products]> search-topk 0.1,0.1,0.1,0.1 10 --where user_id=alice AND price<10
[products]> search-topk 0.1,0.1,0.1,0.1 10 --where (user_id=alice OR user_id=bob) AND category=books
[products]> search-topk 0.1,0.1,0.1,0.1 10 --where year BETWEEN 2021 AND 2023
[products]> search-topk 0.1,0.1,0.1,0.1 10 --where category IN (books,toys) AND NOT user_id=bob
[products]> range-search 0.1,0.1,0.1,0.1 1.0 --where user_id=alice

--where expression grammar (keywords are case-insensitive):

• Comparison: field = value, field != value, field > value, >=, <, <=
• Set membership: field IN (v1, v2, ...)
• Numeric range: field BETWEEN low AND high
• Boolean composition: AND, OR, NOT, and ( ... ) for nesting
• Values: bare words and quoted strings ('x' / "x") are strings; numbers are numeric. Comparison/range operators require numeric (int/float) fields.

Notes:

• --metadata-fields, --meta, and --where are only available on vector spaces created with --index-type Flat.
• The --metadata-fields and --meta lists must not contain spaces (they are comma-separated).
• --compression TurboQuant2Bits, TurboQuant3Bits, or TurboQuant4Bits is optional on those same Flat+metadata spaces. It stores quantized vectors in memory and on disk (approximately 16× / 10× / 8× smaller than float32, plus a per-vector norm). Search reconstructs approximate float32 values. 6-bit and 8-bit schemes are not available; the TurboQuant library only supports 2/3/4-bit quantization.

See Vector Engine — Metadata Filtering for the full grammar, internals, and error reference.

If you prefer using Homebrew on macOS, you can install ShibuDb directly from our tap:

brew tap shibudb.org/shibudb

# Install ShibuDb
brew install shibudb

# If you already have an older version installed, you can upgrade
brew link shibudb

Use pre-built packages for current Linux distributions. For older distributions, use the source installer below so the ShibuDB binary is compiled against the local system libraries.

macOS (Apple Silicon):

sudo installer -pkg shibudb-{version}-apple_silicon.pkg -target /

Linux (Debian/Ubuntu):

# AMD64
sudo dpkg -i shibudb_{version}_amd64.deb

# ARM64
sudo dpkg -i shibudb_{version}_arm64.deb

Linux (RHEL/CentOS):

# AMD64
sudo rpm -i shibudb-{version}-1.x86_64.rpm

# ARM64
sudo rpm -i shibudb-{version}-1.aarch64.rpm

Use this option on older Linux distributions or when a pre-built package is not compatible with your system. The installer downloads the selected ShibuDB source release, uses the bundled FAISS libraries from resources/, builds shibudb locally, installs the binary under /usr/local/bin, installs FAISS libraries under /usr/local/lib, and runs ldconfig.

# Install the latest release
curl -fsSL https://raw.githubusercontent.com/shibudb-org/shibudb-server/main/scripts/install-linux.sh | bash

# Install a specific release
curl -fsSL https://raw.githubusercontent.com/shibudb-org/shibudb-server/main/scripts/install-linux.sh | bash -s -- --version {version}

The source installer supports Linux amd64 and arm64. It installs build dependencies with apt, dnf, or yum, and downloads a temporary Go toolchain if Go 1.24 or later is not already available.

• User Sessions: Store session data with automatic expiration
• Configuration Management: Application and system configuration
• Caching Layer: High-performance caching for applications
• Feature Flags: Dynamic feature toggles and A/B testing

• Recommendation Systems: User and product recommendations
• Image Search: Similar image retrieval and classification
• Text Similarity: Document search and semantic matching
• Anomaly Detection: Identify unusual patterns in data
• Face Recognition: Biometric authentication systems

• SaaS Platforms: Isolated data per customer
• Microservices: Service-specific data storage
• Analytics: Separate spaces for different data types

# Start server (default listen port 4444; management API on 5444)
sudo shibudb start

# Or choose ports (client and management must differ)
sudo shibudb start --port 9090 --management-port 19090

# Stop server
sudo shibudb stop

shibudb manager --username admin --password admin generate-token

# Check status (admin-only; default management port 5444)
shibudb manager --username admin --password admin status
shibudb manager --port 19090 --username admin --password admin status
# Check status (default management port 5444)
shibudb manager status
shibudb manager --port 19090 status

# View connection statistics (admin-only)
shibudb manager --username admin --password admin stats

# Update connection limit (admin-only)
shibudb manager --username admin --password admin limit 2000

# Health check (admin-only)
shibudb manager --username admin --password admin health

# If the server used e.g. --management-port 19090
shibudb manager --port 19090 --username admin --password admin stats

# Dump entire database to JSONL format (server must be stopped)
shibudb dump --output backup.jsonl

# Dump a specific space only
shibudb dump --space my_space --output my_space.jsonl

# Restore database (full restore, overwriting existing data)
shibudb restore --input backup.jsonl

# Restore and merge with existing data
shibudb restore --input backup.jsonl --mode merge

# Get connection status (default management port 5444)
shibudb manager --username admin --password admin generate-token

# Get connection status (default management port 5444; auth required)
curl http://localhost:5444/limit \
  -H "Authorization: Bearer <management_token>"
# Update connection limit (auth required)
curl -X PUT http://localhost:5444/limit \
  -H "Authorization: Bearer <management_token>" \
  -H "Content-Type: application/json" \
  -d '{"limit": 2000}'

We welcome contributions! Please see our Contributing Guidelines for details.

1. Fork the repository
2. Create a feature branch: git checkout -b feature/amazing-feature
3. Make your changes and add tests
4. Run the test suite: make test
5. Commit your changes: git commit -m 'Add amazing feature'
6. Push to the branch: git push origin feature/amazing-feature
7. Open a Pull Request

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0) - see the LICENSE file for details.

Important Note: This license requires that if you run a modified version of this software on a network server, you must make the source code available to users of that server. This prevents commercial SaaS providers from using this software without open-sourcing their service.

• Documentation: Wiki
• Issues: GitHub Issues
• Discussions: GitHub Discussions

• FAISS - Vector similarity search
• TurboQuant - Data-oblivious vector quantization (2/3/4-bit)
• Go B-tree - B-tree implementation
• Go Crypto - Cryptographic functions

ShibuDb - Fast, reliable, and scalable database with vector search capabilities.

For detailed information about specific features, please refer to the documentation.