MindsDB Product Updates - November 2025

MindsDB had a packed month of improvements across Knowledge Bases, integrations, SQL operations, documentation, and the OS GUI. This release focuses on making MindsDB faster, more reliable, and more flexible for real-world AI workflows. From major upgrades to semantic search and data ingestion, to new integration capabilities, to smoother SQL and a refreshed onboarding experience, we’ve continued shaping MindsDB into the most powerful platform for AI-native data access and automation.

Whether you're building intelligent applications, managing high-volume enterprise workloads, or experimenting with multimodal agents, this month's updates bring meaningful speed, stability, and usability enhancements across the board.

Knowledge Bases

We’ve improved how data is inserted into Knowledge Bases by making the maximum batch size apply globally, which results in faster and more efficient data loading. Several enhancements were also made to Knowledge Base handling, including fixing integration tests, resolving issues with OpenAI handlers, improving how metadata columns are processed, and adding more reliable filtering options—such as support for GREATER THAN filters and chunk_content filtering for more precise searches.

You can now disable GUI auto-updates, which also removes unnecessary file handlers. We also fixed and improved Knowledge Base functionality, including major updates to SQL behavior and Azure support.

Integrations

MindsDB has improved integrations across several major platforms, including GitLab, Oracle, MSSQL, MySQL,Gong and more. GitLab and Oracle handlers now support custom URLs, and the MSSQL handler includes a new ODBC option for added flexibility. We also resolved issues affecting Zendesk, Groq, and Jira integrations to ensure smoother, more reliable performance.

Support for using Snowflake’s embedding models inside Knowledge Bases has been added as well, giving users more options for powering semantic search. Additionally, the HubSpot handler received multiple fixes—covering CRUD operations, catalog errors, and overall stability—making the integration more robust and dependable for teams using HubSpot with MindsDB. We added SSL verification controls to the ClickHouse handler, increased the MySQL API connection queue for better performance, and delivered several improvements to our Gong integration.

Bug Fixes and Improvements

Several reliability improvements have been made across the platform, including a fix for CI tests on release branches, improved session-based HTTP authentication, and better handling of SELECT queries in mixed-case databases. We also updated the glob dependency from version 10.4.5 to 10.5.0, which boosts overall performance and strengthens security.

In addition to these enhancements, we resolved a number of smaller bugs and vulnerabilities. These fixes include addressing recursive logprobs when a model isn’t found, patching a memory leak, correcting an issue in the Microsoft Access handler, and preventing unnecessary creation of the files handler. Logging has been improved for clearer debugging, batching for Knowledge Bases is now more efficient, caching has been refined, and startup warnings have been significantly reduced—resulting in a smoother, more reliable user experience.

We shipped multiple important fixes, including improving JSON dumping for NumPy arrays, stabilizing MySQL API tests, and strengthening error handling for A2A agent streaming. We also removed unused manual fork approvals, added clearer debug logs, and reduced start-up warnings for a smoother experience.

Plus, we introduced new capabilities—like support for Snowflake embedding models inside Knowledge Bases.

SQL Operations

MindsDB made SQL smoother and easier to work with—fixing connection errors, updating workflows, improving Docker tags, and polishing overall SQL handling. We also delivered a round of integration improvements for Oracle, Snowflake, MySQL, MSSQL, Databricks, and PostgreSQL.

Plus, Microsoft Access, HubSpot, and BigCommerce connectors were validated and updated with important bug fixes. And finally, we improved left outer join rendering and fixed issues with the LAST function when used in joins.

Documentation

The Minds-Castle documentation has been refreshed to meet newer standards, making the instructions clearer, more intuitive, and easier to follow. We’ve also made several improvements across the broader project documentation to keep everything accurate and up to date. This includes expanded explanations for Knowledge Base chunk_content query options, metadata column listings, updates to the Oracle connector, clearer guidance on HTTP authentication, and updated guidance on using the ALTER KNOWLEDGE_BASE command. 

We updated the Oracle Handler README and added automatic checks to confirm connections when databases are modified. Additionally, various typos and unclear sections were cleaned up to provide a smoother reading and onboarding experience for all users.

OS GUI

We expanded the onboarding flow to include more model providers like Anthropic, Amazon Bedrock, Nvidia NIM, and others.

HubSpot is now available as a new integration, too.

We also made several UX improvements—cleaner sidebar links, better handling of long table names—and fixed bugs such as broken cloud logins and incorrect tab content when switching between multiple environments.

Conclusion

November brought meaningful advancements across the entire MindsDB ecosystem — making AI-powered analytics and automation easier, faster, and more dependable for teams of all sizes. With improvements to Knowledge Bases, stronger integrations (including Snowflake embeddings and HubSpot fixes), major SQL polish, refreshed documentation, and an upgraded GUI onboarding flow, MindsDB continues to evolve into a seamless, production-ready platform for building AI that works directly with your data.

As we head into the next release cycle, we’re doubling down on reliability, developer experience, and deeper support for enterprise AI use cases. Stay tuned — even bigger updates are on the way. Check out the full Release Notes here.