Daniel Usvyat

Update model name and base URL in minds Text2SQL example

1

Updated the model name from 'gpt-4' to 'gpt-4o' and changed the base URL for the MindsDB LiteLLM proxy. Also removed the model parameter from the instantiation of the database mind.

## Description Add LLM Based generations endpoint to knowledge base, this will allow standalone RAG on Knowledge base. I have made retrieval_config optional, so if not specified defaults will be...

Add summarization support to Knowledge Base

1

## Description It updates the knowledge base controller to handle summarization parameters, integrates summarization utilities, and modifies the database schema to store summarization configurations. These changes enable summarizing content before...

A2A server

1

## Description Decouple crewai agent from a2a server (#10774) Run instructions: `python -m mindsdb --api=mysql,mcp,http,a2a` example `config.json` (not mandatory as defaults are using if not given) ``` { "a2a": {...

Add JSON chunking support for KB

2

## Description - Introducing a new JSONChunkingPreprocessor to handle JSON data structures during preprocessing. - Updating the preprocessor configuration to include a JSON chunking type and its associated configuration. -...

## Description Introduced methods for evaluating knowledge bases using test queries, calculating metrics, and storing results in the database. Added API endpoints to trigger evaluation, fetch evaluation history, and retrieve...

Add batching to kb insert and embeddings

1

## Description **Knowledge Base Batch Processing Optimization** - Batches kb inserts with a sensible default (100 rows) as well as batches langchain_embeddings - Before: it only inserted embeddings into kb...

## Description # Document Retrieval Tool A flexible document retrieval and question-answering tool that integrates with MindsDB and LangChain. Here's a breakdown of its key features: ## Configuration - Default...

## Description Updated the release configuration to include A2A API with its own host and port. Adjusted the Dockerfile to expose the new A2A port and modified the entry point...

Set default prompt template and log its usage

1

## Description When `mode` is not 'retrieval', assign a default prompt template instead of raising an error. Added a log to inform about the prompt template being used. This ensures...