Great question, @ukuina! We will need some time to do a detailed analysis of MindsDB. But here are some key differences a first glance:

1. First-class support for unstructured data: EVA natively supports querying over videos, audio, etc. For instance, it manages a video as a table, mapping each tuple to a video frame.

2. Cascades-style query optimizer: EVA has an AI-centric query optimizer with built-in support for cost-based query optimization, predicate reordering, caching of expensive models, etc.

3. First-class support for vector embeddings: EVA supports similarity search using embeddings stored in a vector database.

Similarities to MindsDB:

1. EVA is implemented in Python and uses sqlalchemy for connecting to a structured database system (PostgreSQL, MySQL, etc.)

2. EVA supports Huggingface and OpenAI pipelines in the form of user-defined functions.

We will love to hear from the developers of MindDB if our high-level analysis is correct :)