Documentation · Public beta

Build a reliable Nexir deployment.

Nexir is durable, replicated state with Redis-compatible access: a Raft-replicated key/value store in which every acknowledged write is on a majority of nodes, reads are linearizable by default, and failures are reported by name. These pages are written for the people who deploy and operate it and for developers writing against it. Every statement about behaviour is backed by the code or by a recorded run; where something is not yet implemented or not yet measured, the page says so.

Before you beginNexir is useful for evaluation and development, but its command surface and storage format may change. Use beta clusters only with data you can replace. See the release notes for what changed in the current release.

What Nexir provides

  • RESP2 access from redis-cli and standard Redis clients.
  • A deliberately small scalar command surface: GET, SET with its conditional and expiry options (NX/XX/GET/KEEPTTL, EX/PX/EXAT/PXAT), DEL, the counters INCR/DECR/INCRBY/DECRBY, the EXPIRE/TTL/PERSIST family, plus connection and cluster administration commands.
  • Single-node operation and explicitly managed replicated clusters, including snapshot-based catch-up for a node that fell behind and offline recovery from the loss of a majority.
  • Quorum-confirmed writes and linearizable reads by default.
  • Optional client mTLS and Raft peer mTLS, in separate trust domains.
  • Prometheus metrics for Raft, storage, snapshots, garbage collection, disk footprint and TLS; structured process logs; a bounded shutdown with a meaningful exit status.

Choose a path

Try Nexir locally

Install, initialize one node, and write your first key.

Plan a deployment

Choose storage, addresses, topology, and secure network boundaries.

Check compatibility

Review commands, data types, TTL, errors, and consistency.

Operate a cluster

Monitor health, manage membership, and understand recovery.

What Nexir is not

Nexir is not a complete drop-in Redis replacement, an in-memory cache, a sharded Redis Cluster implementation, a vector or semantic memory, or a managed service. It currently has no passwords or ACLs, no key iteration (SCAN), no backup or restore tooling, no migration between storage-format generations and no automated cluster orchestration. The limitations page lists each with what to do instead.

Next
Quick start →Review
Public-beta limitations →