Sammlung von Newsfeeds

Over the past few months or so, we've been building the pgEdge Postgres MCP Server, an open source tool that lets LLMs talk directly to PostgreSQL databases through the Model Context Protocol. It supports Claude, GPT, local models via Ollama, and pretty much any MCP-compatible client you can throw at it.

We continue our series of articles reviewing changes in PostgreSQL 19. This time we'll look at what emerged from the September 2025 CommitFest.

The highlights from the first July CommitFest are available here: 2025-07.

Just a quick post to note the release of pg_clickhouse v0.1.4. This v0.1 maintenance release can be upgraded in-place and requires no ALTER EXTENSION UPDATE command; as soon as sessions reload the shared library they’ll be good to go.

Thanks in part to reports from attentive users, v0.1.4’s most significant changes improve the following:

We’re launching a new series of articles covering the changes coming up in PostgreSQL 19. This first article focuses on the events from last summer’s July CommitFest.

There’s a special kind of joy in watching a database do something heavy… without making your app threads cry.

That’s the promise of pg_background: execute SQL asynchronously in background worker processes inside PostgreSQL, so your client session can move on—while the work runs in its own transaction. 

It’s a deceptively simple superpower:

Part 2 of the MERISE series New here? Part 1 covers why MERISE exists and how it differs from ER modeling. Short version: it’s a conversation framework that forces business stakeholders to articulate their actual rules before you write a single line of SQL. Last time, I introduced MERISE and why this French methodology kicks ass for database modeling. Today, we’re diving into the first and most crucial step: the Conceptual Data Model (MCD).

Over the last few years, one idea has been quietly taking shape in the Postgres world:

Autonomy in Postgres isn’t a feature.

It’s a posture.

It’s not a checkbox you tick in a product matrix.

It’s not a single “autonomous mode” toggle in a control panel.

FOSSASIA PGDay 2026 ​​talk selection committee met to finalize the schedule

Call for Paper Committee - Members:

• Tatsuo Ishii
• Amit Kapila
• Amit Langote

Seattle Meetup took place on January 12 2026, organized by Lloyd Albin and Jeremy Schneider. Claire Giordano delivered a talk there.

My personal blog here is dedicated to tech geek material, mostly about databases like postgres. I don’t get political, but at the moment I’m so irritated that I’m making the extraordinary exception to veer into the territory of flame-war opinionating…

What a great follow-up for our second PostgresEDI meetup! 🐘

First off, a huge thank you for braving the snow 🌨️ last evening in Edinburgh , and many thanks to the two speakers who made it a great night.

A lot of the time in database land, our queries are I/O constrained. As such, performance work often involves reducing the number of page reads. Indexes are a prime example, but they don’t solve every issue (a couple of which we’ll now explore).

Securing sensitive data requires more than just a firewall; let's see how Oracle implements VPD and PostgreSQL implement Row and Column level security

How ParadeDB uses principles from search engines to optimize Postgres' Top K performance.

Bluebox Docker provides a realistic, continuously-updating sample database for PostgreSQL users. It automates data generation, simulating a video rental kiosk with real movie data. Users can easily set it up with a single command, facilitating learning and practice of PostgreSQL features without dealing with stale data.

Only ONE WEEK LEFT for the earlybird tickets. Make sure you get yours before it's too late!

https://2026.pgday.paris/registration/

This was a long requested server side feature that have been committed recently to PostgreSQL core to inform users that their passwords must be changed before the expiration limit.

A new entity in the PostgreSQL landscape.

Introducing mnemosyne systems

The last week a new entity appeared in the PostgreSQL landscape: **mnemosyne systems **, the last creation of Jesper Pedersen.

The initial page says it all:

The short answer is always “maybe”. However, in the following post, I hope to demonstrate what creates a sub-transactions and what happens to the overall transaction id utilization when they are invoked. I will also show how performance is affected when there are lots of connections creating and consuming sub-transactions.

First, it is important to understand what statements will utilize a transaction id and which ones may be more critical (expensive) than others:

The PostgreSQL ecosystem just introduced a new pg_stat_plans extension. It's similar to pg_stat_statements but it tracks aggregated statistics for query plans instead of SQL statements.

Large language models know a remarkable amount about PostgreSQL. They can write SQL, explain query plans, and discuss the finer points of MVCC with genuine competence. But there are hard limits to what any model can know, and when you're building tools that connect LLMs to real databases, those limits become apparent surprisingly quickly.The core issue is training data. Models learn from whatever was available at the time they were trained, and that corpus is frozen the moment training ends.