Bringing LLMs to SQL: Exploring ClickHouse's New AI Functions.
Article excerpt
Bringing LLMs to SQL: Exploring ClickHouse's New AI Functions. June 3, 2026 Introduction. Modern data systems are evolving fast and the line between analytics and artificial intelligence is starting to blur. Traditionally, if you wanted to apply AI models like Large Language Models (LLMs) to your data, you had to build separate pipelines, integrate APIs, and manage complex workflows. However, with the latest updates in ClickHouse, that's changing. ClickHouse introduces native AI functions that allow you to directly interact with LLMs using SQL. As a result, you can now run AI-powered queries on your data in real time without leaving your database. In this blog, Quantrail Data private Ltd will walk through what those functions are, how to set them up, and how to use them correctly. A quick note: these are experimental. These AI Functions were officially introduced in experimental releases around ClickHouse version 25.x+, with availability depending on environment ( Cloud or Specific builds). Additionally Since they are part of a rapidly evolving AI suite, the documentation focuses primarily on the shared infrastructure like retries and token management rather than deep per-function guides. If you find these functions aren't enabled in your version, run this first: SET allow_experimental_ai_functions = 1; NOTE: This flag enables the feature only if your build...
Keep reading with a free account
The rest of this article, and every signal for ClickHouse, is in your free account.
Extracted from this sentence
With the recent release, ClickHouse integrates directly with leading AI providers like OpenAI and Anthropic.
