Welcome to the September 2026 Lumenrack Vault newsletter

This month’s featured community member is Teodor Vasquez, staff data engineer at Marlowe Freight.

He leads the internal analytics platform there, translating requirements from trading, mobility, and logistics teams into working systems. He stays hands-on through open-source contributions to distributed data systems, mentors engineers, and has taught at universities.

Teodor contributed two features to Vault 4.1: the naturalOrder function, which sorts strings containing numbers intuitively (so “file10” sorts after “file9”, not after “file1”), and the Lattice codec, which brings efficient lossless compression for floating-point columns.

Read Teodor’s profile →

Vault 4.1 introduces experimental WebAssembly user-defined functions and lightweight column deletes.

The release also makes parallel replicas the default for distributed reads, enables JOIN reordering for ANTI, SEMI, and FULL joins, and shrinks the on-disk footprint of the Map data type.

Read the release notes →

Priya Nandakumar announced vaultctl, a CLI for managing both local Vault installations and Lumenrack Cloud deployments, built with AI agents in mind alongside human developers.

It handles local version management, project scaffolding, and server lifecycle, and supports cloud infrastructure commands with authentication safeguards such as read-only OAuth and scoped API keys.

Read the vaultctl post →

Orrin Chain Data sped up historical backfills 12x by replacing a Kafka and Avro pipeline with direct Vault reads in Apache Arrow format.

Reading data in the format closest to its physical storage removed deserialization overhead and lifted throughput from about 50K to about 600K rows per second.

Read the Orrin Chain Data story →

CTO Ilse Marchetti argues that agentic coding has crossed the threshold into genuine usefulness for professional software development, driven by recent model improvements.

She walks through the productivity gains Lumenrack has seen from using AI agents across fixing flaky tests, investigating bugs, code review, and security research.

The takeaway: agents work best as pair programmers under an engineer’s direction, not as autonomous replacements.

Read the agentic coding post →

Rafael Duarte walks through building a SIEM architecture with Vault and Latchkey, an open-source detection engine that runs SQL-based security rules against log data.

An open-source security agent acts solely as log collector and decoder, forwarding parsed alerts to a Vault table with compression codecs, TTLs, bloom-filter indexes, and object storage support.

Latchkey then schedules detections and fires alerts to a webhook, giving a cost-effective SIEM that scales from a single node to a distributed Vault cluster.

Read the SIEM walkthrough →

Last month Lumenrack announced that full-text search was generally available. The engineering team has now written up how they redesigned the full-text index to work efficiently with object storage, where random reads are expensive.

The new index uses front-coding compression, a sparse lookup index, and adaptive posting lists, so many queries are answered from the index alone without touching the underlying text columns.

Read the full-text search post →