Weaviate 1.37 Release

Weaviate v1.37 ships several major features. A built-in MCP Server (preview) enables LLMs and AI agents to interact with Weaviate natively via Claude Code, Cursor, VS Code, and other MCP-compatible clients — exposing tools for schema inspection, hybrid search, and data upsert. Extensible Tokenizers (preview) add accent folding, per-property custom stopword presets, and a /v1/tokenize endpoint for debugging. Diversity Search via Maximum Marginal Relevance (MMR, preview) reranks vector results at query time to reduce near-duplicate clustering. Query Profiling (preview) provides per-shard timing breakdowns for any search request. Incremental Backups reduce backup size and time by referencing unchanged file chunks from a base backup. The multi2vec-google module gains audio as a fourth modality (Gemini API only). A new BlobHash property type stores only a SHA-256 hash after vectorization, saving disk space. Additional improvements include collection export to Parquet, gRPC replica communication, IPv6 clustering, Cohere reranker v2, and numerous performance fixes.