- Added experimental AIâpowered search with vector store, GPUâaccelerated HuggingFace embeddings, and new OpenAI embedding models (textâembeddingâ3âsmall/large) with configurable dimensions.
- Improved indexing speed and reduced memory usage via autoâbatch task deletion, hybrid search speedâup, and capped sorter memory; showRankingScoreDetails is now stable.
- Introduced experimental JSON logging and new /logs/stream and /logs/stderr routes, plus various bug fixes and documentation enhancements.