github.com/NodeDB-Lab/nodedb @v0.3.0

The distributed multi-model database for AI and agent workloads.

Eight database engines in a single Rust binary. One SQL dialect. Zero network hops between vector search, graph traversal, document storage, columnar analytics, timeseries, key-value, full-text search, and multi-dimensional arrays.

Quickstart · Docs · Benchmarks · NodeDB-Lite · CLI

NodeDB replaces the combination of PostgreSQL + pgvector + Redis + Neo4j + ClickHouse + Elasticsearch with a single process. Graph queries that feed vector search, full-text ranking, and columnar aggregation execute in one engine with shared storage, shared memory, and one planner.

Why NodeDB

• One binary, not a polyglot stack. No inter-service networking, no schema drift between systems, no data synchronization pipelines. A graph query that feeds a vector search that filters by full-text relevance executes in one process.
• PostgreSQL wire protocol. Connect with psql or any PostgreSQL client library. Standard SQL with engine-specific extensions where SQL can't express the operation.
• Edge to cloud. The same engines run embedded on phones and browsers (NodeDB-Lite, WASM) with CRDT-based offline-first sync to the server.
• Serious about performance. Thread-per-Core data plane with io_uring, SIMD-accelerated distance functions, zero-copy MessagePack transport, per-column compression (ALP, FastLanes, FSST, Gorilla). See benchmarks below.

Performance

Timeseries ingest + query benchmark — 10M rows, high-cardinality DNS telemetry (50K+ unique domain names). Single node, NVMe storage.

Ingest

Queries (ms, best of 3)

NodeDB is not a specialized timeseries database, yet it ingests 1.65x faster than TimescaleDB and 1.73x faster than ClickHouse with 8x less memory. Query latency is competitive with ClickHouse on low-cardinality aggregations and within 3-5x on high-cardinality GROUP BY. This is the tradeoff of a general-purpose engine: you get one system instead of five, with performance that stays in the same ballpark as specialized tools.

Engines

Install

# Docker
docker run -d \
  -p 6432:6432 -p 6433:6433 -p 6480:6480 \
  -v nodedb-data:/var/lib/nodedb \
  farhansyah/nodedb:latest

# Cargo
cargo install nodedb

Requires Linux kernel >= 5.1 (io_uring). Connect:

ndb                              # native CLI (connects to localhost:6433)
psql -h localhost -p 6432        # or any PostgreSQL client

CREATE COLLECTION users;

-- Standard SQL
INSERT INTO users (name, email, age) VALUES ('Alice', 'alice@example.com', 30);

-- Object literal syntax (same result)
INSERT INTO users { name: 'Bob', email: 'bob@example.com', age: 25 };

-- Batch insert
INSERT INTO users [
    { name: 'Charlie', email: 'charlie@example.com', age: 35 },
    { name: 'Dana', email: 'dana@example.com', age: 28 }
];

SELECT * FROM users WHERE age > 25;

New here? Start with the Quickstart on the official documentation site: nodedb.dev/docs.

Deployment Modes

NodeDB-Lite

All eight engines as an embedded library. Linux, macOS, Windows, Android, iOS, and browser (WASM, experimental).

• Lite only -- local-first apps that don't need a server. Vector search, graph, FTS, documents, arrays, all in-process with sub-ms reads.
• Lite + Origin -- offline-first with CRDT sync. Writes happen locally, deltas merge to Origin when online. Multiple devices converge regardless of order.
• Same API -- the NodeDb trait is identical across Lite and Origin. Switch between embedded and server without changing application code.

See NodeDB-Lite for platform details and sync configuration.

Key Features

Write-time validation -- Typeguards enforce types, required fields, CHECK constraints, and DEFAULT/VALUE expressions on schemaless collections. Graduate to strict schema with CONVERT COLLECTION x TO document_strict.

Bitemporal queries -- System time (audit trail) and valid time (temporal semantics) across all engines. Query data as it existed in the past, or as it was valid on a past date. GDPR-compliant tile purge on array engine.

Multi-dimensional arrays -- Scientific computing with Z-order indexed tiles, tile-level compression, and bitemporal support. Combine with vector/graph/text in fused queries via cross-engine identity.

Real-time -- CDC change streams with consumer groups (~1-5ms latency). Streaming materialized views. Durable topics. Cron scheduler. LISTEN/NOTIFY. All powered by the Event Plane.

Programmability -- Stored procedures with IF/FOR/WHILE/LOOP. User-defined functions. Triggers (async, sync, deferred). SECURITY DEFINER.

Security -- RBAC with GRANT/REVOKE. Row-level security with $auth.* context across all engines. Hash-chained audit log. Multi-tenancy with per-tenant encryption. JWKS, mTLS, API keys.

Six wire protocols -- pgwire (PostgreSQL), HTTP/REST, WebSocket, RESP (Redis), ILP (InfluxDB line protocol), native MessagePack.

Tools

• ndb -- Native CLI with TUI, syntax highlighting, and tab completion. Alternative to psql.
• NodeDB Studio -- GUI client for managing collections, browsing data, and monitoring. (coming soon)
• nodedb-bench -- Performance benchmarks against competing databases.

Documentation

The official documentation site is nodedb.dev/docs — start with the Quickstart.

In-repo references:

• Getting Started -- Build, run, connect
• Architecture -- Three-plane execution model
• Engine Guides -- Deep dives into each engine
• Security -- Auth, RBAC, RLS, audit, multi-tenancy
• Real-Time -- CDC, pub/sub, LIVE SELECT
• NodeDB-Lite -- Embedded edge database
• AI Patterns -- RAG, GraphRAG, agent memory, feature store

Contributing

We welcome bug fixes, engine improvements, new codecs and analyzers, test coverage, and documentation. Read CONTRIBUTING.md before opening a PR — NodeDB's Three-Plane execution model has hard rules that reviewers enforce.

Building from Source

For development or contributing:

git clone https://github.com/NodeDB-Lab/nodedb.git
cd nodedb
cargo build --release
cargo install cargo-nextest --locked  # one-time
cargo nextest run --all-features

Release Status

NodeDB Origin is in public beta as of v0.1.0 (2026-05-07). All eight engines are feature-complete and covered by tests. The wire protocols (pgwire, HTTP, native MessagePack, RESP, ILP, WebSocket) are stable — clients written against 0.1.0 will keep working through 1.0.

v0.1.0 — Beta (today). Build new products on it. The public surface (SQL dialect, wire protocols, configuration) is stable; expect internal changes (storage layout, on-disk format, replication internals) between minor releases. Patch and minor bumps will land as needed — if something requires a 0.2, we ship a 0.2. The two months between beta and 1.0 are deliberately for real workloads to surface edge cases we can't manufacture in-house.

v1.0.0 — Production-ready (target: 2026-07-07). What 1.0 guarantees:

• API & SQL stability — semver from 1.0 onward. No breaking SQL or client-API changes within a major.
• Wire protocol stability — pgwire, HTTP, native MessagePack, RESP, ILP, WebSocket frozen.
• On-disk format stability — no breaking migrations within 1.x. Forward-compatible upgrades only.
• Cluster & Raft stability — rolling upgrades supported within 1.x; no quorum-breaking changes.
• Performance SLAs — published p50/p99 targets per engine, regression-gated in CI.
• Security audit — third-party audit completed and findings remediated before 1.0 ships.
• Storage, backup, and recovery — fully exercised under fault injection, sustained load, and crash-restart cycles.

Pre-1.0 versions may change internals between releases — those changes are critical-path work (storage, backup, security, recovery) that has to be hardened in real production conditions before we put a sta