Weaviate v1.38 is now available open-source and on Weaviate Cloud. Two features reach general availability: HFresh, a disk-based vector index inspired by the SPFresh algorithm that keeps memory low for streaming workloads with built-in RQ-1 quantization, and the built-in MCP Server at /v1/mcp enabling LLMs and AI agents to interact with Weaviate directly. Async replication has been rebuilt to run cluster-wide from a single scheduler and is now on by default for all replicated collections. Two previews are introduced: the Boost API for query-time rescoring that promotes or demotes results without dropping any, and Nested Object Filtering using dotted paths for object/object[] properties. Additional improvements include production-ready replica movement, configurable default vector index types, usage guardrails, a new text2vec-digitalocean module, fractional BM25 property boosts, and deterministic tie-breaking in vector searches.
