TL;DR

Bug Fixes and Improvements

• The output of Evaluate has been fixed, by removing retrieved_in_top_k, which helps to make results cleaner.
• Invalid model error with Azure openai models has been ironed out to allow smoother execution.
• Successful completion is now a requirement for test setup, improving the reliability of tests.
• The deletion issue within ML engine with data catalog enabled has been resolved.
• Various fixes have been implemented, like correcting the 'SELECT' from API Resources, fixing the querying of Snowflake columns with special characters, and fixing the ssl option of MySQL handler. These contribute to a smoother, more reliable user experience.

Changes to SQL

• An adjustment has been made to reduce token usage in generating schema information for agents, which can lead to performance improvements.

Changes to Integrations

• Google BigQuery is now a supported source in Data Catalog, expanding the usability of the platform.
• Fixes to 'Agent Salesforce' have been introduced to enhance functionality.
• The llama_index_handler has been updated from version 0.12.21 to 0.12.28, ensuring that the handler operates using the latest available features and fixes.

Changes to Knowledge Bases

• Knowledge base documentation has gone through numerous updates, for instance, 'evaluate kb', 'evaluate syntax', and new links to AI integrations have been added. This increases the resourcefulness of the platform.
• Knowledge base for Python SDK has been hidden.
• The REST API Endpoints for Knowledge Bases has been updated for better accessibility and usability.
• The platform now provides functionality for enhanced conversation history handling in MindsDBAgent and AgentTaskManager, creating a more efficient system.

Changes to Documentation

• Documentation has been updated with usage examples in 'ollama knowledge base', creating a more user-friendly guide.

What's Changed

• Fix Evaluate output: remove retrieved_in_top_k by @ea-rus in #11050
• Reduced Token Usage in Generating Schema Information for Agents by @MinuraPunchihewa in #11049
• remove prod slack msg by @hamishfagg in #11047
• Update constants.py by @ZoranPandovski in #11045
• fix invalid model error on azure openai models by @fshabashev in #11075
• hide kb docs of python sdk by @martyna-mindsdb in #11078
• kb docs -- evaluate kb by @martyna-mindsdb in #11035
• Require success for test setup by @hamishfagg in #11051
• kb docs -- evaluate syntax by @martyna-mindsdb in #11080
• Fix ML engine deletion with data catalog enabled by @hamishfagg in #11082
• Fixed SELECT from API Resources by @MinuraPunchihewa in #11084
• Updated the REST API Endpoints for Knowledge Bases by @MinuraPunchihewa in #10813
• Enabled Google BigQuery as a Supported Source in Data Catalog by @MinuraPunchihewa in #10947
• A2A: Enhance conversation history handling in MindsDBAgent and AgentTaskManager by @setohe0909 in #11083
• Bump llama-index from 0.12.21 to 0.12.28 in /mindsdb/integrations/handlers/llama_index_handler by @dependabot in #11032
• update documentation with ollama knowledge base usage example by @fshabashev in #11103
• Agent Salesforce fixes by @ZoranPandovski in #11077
• Fixed Querying Snowflake Columns with Special Characters by @MinuraPunchihewa in #11101
• Fix ssl option of mysql handler by @StpMax in #11102
• kb docs -- added links to ai integrations by @martyna-mindsdb in #11079

Full Changelog: v25.6.2.0...v25.6.3.0