Introduction

NYXDB is a PSI-based, low-latency HTAP database engine — streaming SQL, keyed tables with exact counts, read-your-writes, and continuous transforms in one runtime.

NYXDB is a distributed, high-throughput HTAP database engine for large-scale blockchain and analytics workloads. It is built for systems where fresh data, fast reads, streaming updates, and programmable computation must live inside one coherent runtime — trading, on-chain analytics, telemetry, and observability.

Start here

Getting started

Run a node with Docker and issue your first queries in minutes.

Core concepts

Table kinds, transforms, streaming reads, storage, and indexes.

SQL reference

Per-statement syntax, data types, and function families.

Operations

Deploy, tune the memory governor, and observe with system tables.

Drivers & protocol

The nyxsql CLI, JDBC, and the wire protocol.

What makes NYXDB different

Most low-latency stacks split one workflow across an OLTP database, an analytical engine, a stream processor, a cache, and a message bus. NYXDB collapses those into one engine where streaming, querying, indexing, durability, and function execution share the same internal contracts:

Append, keyed, and attribute tables — arrival-order history, O(1) latest-per-key current state with exact counts, and point-in-time reconstruction, all in one catalog.
Continuous transforms — CREATE TRANSFORM defines standing pipelines that stay fresh through PSI routing, with gap-free recovery and pause/resume.
Streaming reads — STREAM SELECT follows results live; STREAM HISTORICAL replays then follows.
Read-your-writes freshness — a just-committed write is visible to the next read, with no cache to keep in sync.

NYXDB is an open-core engine in active development. The public runtime proves the single node; the enterprise runtime adds high-availability features behind the same interfaces. Snippets in these docs are verified against the engine's own tests; anything not yet confirmed is marked TODO-verify.