⚠️ Early alpha — please test on a copy and report issues. This is a structural release; it was tested against real upgrade data but not every possible 1.x layout. Back up your database and config before upgrading.
What's new in 2.0
2.0 is a ground-up refactor of how servers and models are managed:
Multi-instance AI provider for llama.cpp and any OpenAI-compatible server. Supports chat, embeddings, speech-to-text, rerank, moderation and text-to-image.
This release expands the provider beyond chat and embeddings, adding two new operation types and automatic capability detection so that models work out of the box with minimal configuration.
