What if you put an LLM inside SQL ?

Join us for another engaging session with Mehdi and Till as we explore the AI features built into Motherduck and the open-source initiatives we've been working on.
No fluff—just a showcase of what's live in production, how we're simplifying SQL workflows, and a discussion about our future plans.

☁️🦆 Start using DuckDB in the Cloud for FREE with MotherDuck : https://hubs.la/Q02QnFR40

00:00 About Till - one foot in the industry, one foot in research
11:38 MotherDuck AI SQL reference
15:39 Generate SQL query or direct insights
19:20 FixIt - AI one line SQL fixer
26:31 About Pricing
28:05 OpenSource DuckDB SQL model
34:30 The prompt() function history
38:10 The prompt() function demo
43:00 Prompt() with structured output
48:29 Embeddings
51:35 What's next

➡️ Follow Us
LinkedIn: / motherduck
X/Twitter : / motherduck
Blog: https://motherduck.com/blog/

--------------------------------------

Explore the cutting-edge integration of AI and Large Language Models (LLMs) directly into the DuckDB and MotherDuck ecosystem. In this deep dive, MotherDuck AI engineer Till joins us to demonstrate how we're moving beyond external chatbots to embed powerful AI features directly into your SQL workflow. We'll cover practical tools designed for data engineers and analysts, showcasing how to generate, debug, and enhance your analytical queries using generative AI, making your data exploration faster and more intuitive.

Discover how to leverage text-to-SQL capabilities with functions like `prompt_sql` to instantly generate schema-aware queries from natural language questions. We then introduce the game-changing "Fix It" feature in the MotherDuck UI, an intelligent assistant that provides immediate, line-specific suggestions to correct syntax and schema errors. This creates an incredibly fast feedback loop for SQL query fixing, a significant improvement over regenerating entire queries. We also discuss the technology behind these tools, including our open-source, fine-tuned DuckDB SQL model that you can run locally with Ollama for enhanced privacy and control.

Unlock advanced NLP capabilities directly within your database using the versatile `prompt()` function. Learn how to perform complex tasks like on-the-fly text summarization, translation, and sentiment analysis. We highlight its most powerful use case: structured data extraction. See how you can command the AI to parse unstructured text columns and return clean, queryable JSON objects, transforming raw data into valuable insights without ever leaving your SQL editor. This is ideal for anyone looking to implement a data pipeline for text analysis using familiar SQL syntax.

Finally, we touch on creating embedding models for building vector search and RAG pipelines entirely within SQL, expanding the possibilities for semantic search on your data. We look to the future, discussing the rise of smaller, specialized language models and the potential for local inference in the browser with WebGPU. This session provides a comprehensive overview of how MotherDuck is integrating AI to augment the developer experience, reduce friction in data analysis, and unlock new ways to work with your data in DuckDB.