2 posts tagged with "LLM"

In one of the recent PostgresFM episodes, Michael Christofides and Nikolay discussed planning time and what can affect it. One of the obvious negative factors we've discussed is the large number of partitions a partition table might have.

In this post, I'm using our Postgres.AI assistant to see how planning time depends on the number of partitions.

This is the very first blog post of its kind: it has an integrated AI that you can use to explore the topic discussed here further, to repeat the experiment, alter it in any direction, and study the Postgres behavior more effectively.

I'm happy to present our new product, Postgres.AI Bot. It is powered by OpenAI's GPT-4 Turbo and is designed to help engineers improve their experience when working with PostgreSQL. This bot has a vast knowledge base with over 110,000 entries, including documentation, source code for different PostgreSQL versions, and related software like PgBouncer, Patroni, and pgvector. It also integrates expert articles and blogs.

In addition, the bot can conduct two types of experiments to verify ideas:

1. Single-session on thin clones provided by Postgres.AI DBLab Engine to check SQL syntax and behavior of PostgreSQL planner and executor, and
2. Full-fledged benchmarks (pgbench) on separate VMs in Google Cloud to study Postgres behavior under various workloads. For each iteration, 70+ artifacts are automatically collected and used by the bot to analyze and visualize experiment results.

Our ambitious goal for 2024 – conduct 1 million database experiments in both shared and dedicated environments, aiding both the bot's knowledge and the improvement of these projects, particularly in performance. To achieve this, we became a part of Google Cloud's AI startup program.

In this blog post, we discuss some details of how the bot is implemented, what it is capable of, share its first exciting achievements, and talk about the future.